A method for parameter identification of heat pump systems based on Adam's improved co-evolutionary algorithm

Abstract

This invention provides a method for identifying parameters of a heat pump system based on an improved Adam co-evolutionary algorithm, belonging to the field of industrial control process system identification technology. It solves the technical problems of traditional algorithms in heat pump system parameter identification, such as lack of physical constraint guidance, easy output of meaningless solutions, slow convergence speed, and the tendency of co-evolutionary algorithms to get trapped in local optima. The method includes the following steps: Step 1) Constructing a Hammerstein-Wiener model of an air-source heat pump based on the thermodynamic mechanism of core components; Step 2) Constructing the identification process of the improved Adam co-evolutionary algorithm, combining three groups of cooperative search, chaotic mechanisms, and the Adam optimizer to achieve efficient and accurate parameter identification. This invention uses the improved Adam co-evolutionary algorithm for air-source heat pump system parameter identification, which has both fast convergence speed and high identification accuracy, effectively avoids local optima, and has small parameter estimation errors.

Description

Technical Field

[0001] This invention relates to the field of heat pump system identification technology, and in particular to a method for identifying heat pump system parameters based on the Adam improved co-evolutionary algorithm. Background Technology

[0002] Air source heat pumps have become a core energy-saving device in the building HVAC field. Building energy consumption accounts for over 40% of my country's total social energy consumption, with HVAC systems accounting for 50% of that. Air source heat pumps, with their "low-energy-consumption-driven reverse Carnot cycle" characteristic, have become a key technology for reducing building energy consumption. However, air source heat pump systems are typical multivariable nonlinear systems. Their core components (compressor, evaporator, condenser, and expansion valve) all exhibit significant nonlinear characteristics. Traditional linear control methods are prone to problems such as decreased control accuracy, significant dynamic overshoot, and delayed steady-state response.

[0003] To achieve precise control of air source heat pump systems, accurate mathematical models and parameter identification are necessary. Block-structured nonlinear systems can effectively describe such nonlinear dynamic systems. The Hammerstein-Wiener model, as an important branch of this model, consists of an input static nonlinear module, an intermediate dynamic linear module, and an output static nonlinear module connected in series, and has been used to characterize air source heat pump systems.

[0004] Existing parameter identification methods have several shortcomings: traditional swarm intelligence algorithms (such as CEA and PSO) lack physical constraints during parameter identification, easily outputting meaningless solutions, and have slow convergence speeds, making it difficult to meet the real-time requirements of engineering projects; existing Hammerstein-Wiener models mostly use pure mathematical function fitting, failing to consider the thermodynamic mechanisms of the core components of heat pumps, resulting in large deviations between model outputs and actual operating characteristics, especially significant errors in heat transfer efficiency prediction under ultra-low temperature conditions. Therefore, providing a parameter identification method that balances identification accuracy, convergence speed, and physical interpretability has become a pressing technical problem to be solved in this field.

[0005] The objective of this invention is to solve the aforementioned technical problems. Summary of the Invention

[0006] The purpose of this invention is to provide a method for identifying parameters of a heat pump system based on the Adam improved co-evolutionary algorithm. By constructing a physically interpretable Hammerstein-Wiener model and combining it with the Adam improved co-evolutionary algorithm, the method achieves efficient and accurate identification of parameters of an air source heat pump system, solving the technical problems of existing methods that are prone to getting trapped in local optima, have low identification accuracy, slow convergence, and unclear physical meaning.

[0007] This invention is achieved through the following measures: a method for parameter identification of an air source heat pump system based on the Adam improved co-evolutionary algorithm, specifically including the following steps:

[0008] Step 1) Construct a Hammerstein-Wiener model of a heat pump system based on the thermodynamic mechanism of its core components to obtain a physically interpretable identification model of the heat pump system;

[0009] Step 2) Construct the identification process for the Adam improved co-evolutionary algorithm.

[0010] As a further optimization scheme for the heat pump system parameter identification method based on the Adam improved co-evolutionary algorithm provided by the present invention, the specific modeling steps of step 1) are as follows:

[0011] Step 1-1) Construct nonlinear models of the core components of the air source heat pump, including a nonlinear model of the compressor, a nonlinear model of the expansion valve, an ambient temperature correction model, and a thermodynamic mechanism model of the hot water outlet temperature:

[0012] Compressor nonlinear model: Based on thermodynamic mechanisms and the ideal gas law, a saturation function is introduced to characterize the physical saturation characteristics of the rotational speed. The model form is as follows:

[0013]

[0014] Where η v Let u1(t) be the volumetric efficiency, u1(t) be the compressor speed input, and V be the volumetric efficiency. th Theoretical displacement, ρ(t) is the refrigerant density. (ρ is pressure, m is molar mass, R is ideal gas constant, T is refrigerant temperature); according to Figure 2 A compressor model with saturation characteristics can be represented as:

[0015]

[0016] Where x1 and x2 are constants.

[0017] Nonlinear model of expansion valve: based on the pressure difference Δp = p across the valve. A -p B (p A p is the condenser outlet pressure. B Modeling the evaporator inlet pressure for both turbulent and laminar flow states, and introducing a pressure drop threshold. The model is integrated as follows:

[0018]

[0019] Where u2(t) is the expansion valve opening input, k val The expansion valve flow coefficient;

[0020] Ambient temperature correction model: An ambient temperature correction term is introduced to compensate for the decrease in the evaporator heat transfer coefficient. The equivalent input is:

[0021]

[0022] Where c1<0 is the ambient temperature correction factor, T a (t) represents the ambient temperature;

[0023] Hot water outlet temperature model: Combining the heat transfer and exponential saturation characteristics of the condenser, the model is as follows:

[0024]

[0025] Where T w，in ΔT represents the inlet temperature of the cold water. max The maximum temperature rise is given by k, which is a nonlinear coefficient, and T is the temperature rise. co This refers to the condenser temperature.

[0026] Steps 1-2) Integrate the compressor nonlinear model, expansion valve nonlinear model, ambient temperature correction model, and hot water outlet temperature thermodynamic mechanism model to construct a heat pump Hammerstein-Wiener model, based on... Figure 1 The system can be represented as:

[0027] x(t)=f1[u1(t)]+f2[u2(t)]+f3[u3(t)] (6)

[0028]

[0029] y(t)=h[w(t)] (8)

[0030] In the formula, t is the time variable, u(t) and y(t) are the system's input and output signals, respectively, and v(t) is the equivalent correction input for ambient temperature. x(t) is the output of the nonlinear part of the system, and w(t) is the linear dynamic part of the system. -1 It is the unit shift operator: q -1 =u(t-1). A(q) -1 B(q) -1 ) is q -1 The constant polynomial in the equation is defined as follows:

[0031]

[0032] Where a i It is a polynomial A(q) -1 The factor of ) b i It is a polynomial B(q) -1The factor of ) and the order n are known. The relationship between the output w(t) and the output x(t) of the nonlinear part can be obtained from formula (6):

[0033]

[0034] Based on equations (1), (3), (4), and (6), they are updated as follows:

[0035]

[0036] According to equation (5), equation (8) is updated to:

[0037]

[0038] Where T w，in ΔT max k1, k2, and c1 are constants.

[0039] Steps 1-4) Define the parameter vector to be identified, integrating linear parameters, nonlinear coefficients, and physical parameters, in the form of:

[0040] θ = [a1, a2, ..., a n b1, b2, ..., b n [k1, k2, c1, Δp] T

[0041] As a further optimization of the heat pump control system identification method based on the Adam improved co-evolutionary algorithm provided by the present invention, the specific process of step 2) includes the following steps:

[0042] Step 2-1) Initialization settings: Given the number of iterations k, population size N, and data length M, define the parameter search space [lb]. j ub j (j is the parameter index); Three groups are divided into pole search groups (PS) θ ps = [a1, a2, ..., a n ] T Zero-point search group (ZS)θ zs = [b1, b2, ..., b n ] T Nonlinear parameter search group (NP)θ np =[k1, k2, c1, Δp] T The initial population is generated based on the Logistic chaotic mapping, and the chaotic sequence is mapped to the parameter space, as shown in the formula:

[0043] z n+1 =4·z n (1-z n), z0∈(0,1), n=1,2,...,N (14)

[0044] Step 2-2) Data Acquisition: Use the compressor speed u1(t), expansion valve opening u2(t), and ambient temperature u3(t) of the air source heat pump as input data, and the hot water outlet temperature y(t) as output data, and record M sets of sample data;

[0045] Step 2-3) Fitness Function Definition: A stability penalty term is introduced to ensure system stability. The fitness function is:

[0046]

[0047] Where y meas (t) represents the measured output, y sim (θ,t) represents the model simulation output, r i For the poles of the CARMA model, n s For the number of extreme points; Steps 2-4) Explore the co-evolution of the three populations: subpopulations evolve independently, and the PS population fixes θ. zs and θ np , for θ ps Perform chaotic mutation (early stage) or Gaussian mutation (late stage); fix θ in the ZS group. ps and θ np , for θ zs Perform similar mutations; NP group fixes θ ps and θ zs , for θ np Perform similar mutations; the mutation formula is: Chaotic Mutation Gaussian mutation

[0048] Steps 2-5) Cross-population elite information fusion: Select the top M best individuals from each of the three populations to form an elite set. Replace the nonlinear parameters of the PS and ZS populations with the elites of the NP population. Replace the zero-point parameters of the PS and NP populations with the elites of the ZS population. Replace the extreme parameters of the ZS and NP populations with the elites of the PS population. After evaluation, retain the better solution.

[0049] Steps 2-6) Population Reconstruction and Global Optimal Update: Merge the three populations, select the top N best individuals by fitness, re-divide them into a new generation of three populations, and update the global optimal solution θ. * ;

[0050] Steps 2-7) Adam Optimization and Refinement: When the population shows no performance improvement for three consecutive generations, start the Adam optimizer to refine the globally optimal parameter θ. * Refining:

[0051] The gradient of the fitness function can be approximated by numerical differentiation:

[0052]

[0053] Where h is the differential step size, e j It is a unit vector.

[0054] Calculating the first-order momentum (momentum term) is used to accumulate historical gradient directions, accelerating convergence:

[0055]

[0056] Where β1 = 0.9 is the momentum decay coefficient, and t is the Adam iteration number.

[0057] Calculate second-order momentum (adaptive learning rate): Adjust the step size based on the squared gradient to suppress oscillations.

[0058] It also compensates for bias, updates parameters using an adaptive learning rate, and refines the θ... * Inject the worst-performing individual into the population;

[0059] Step 2-8) Iteration Termination Judgment: If the maximum number of iterations K is reached or the fitness meets the preset accuracy requirement, output the identification result θ. * Complete system identification.

[0060] To achieve the above-mentioned objectives, this invention also provides a method for identifying parameters of a heat pump system based on the Adam improved co-evolutionary algorithm, the system comprising:

[0061] The physically interpretable heat pump system identification model building module is configured to perform the following process: construct a physically interpretable heat pump system identification model based on the Hammerstein-Wiener model of the heat pump system with the thermodynamic mechanism of the core components;

[0062] The Adam improved co-evolutionary algorithm building module is configured to perform the following process: building the identification process of the Adam improved co-evolutionary algorithm.

[0063] Meanwhile, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed, it implements the steps of the method of the present invention.

[0064] In addition, the present invention also provides a computer-readable storage medium having a computer program stored thereon, the computer program being configured to implement the steps of the method of the present invention when invoked by a processor.

[0065] Finally, the present invention provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the method of the present invention.

[0066] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0067] (1) This invention establishes a high-precision parameter identification model adapted to the strong nonlinear characteristics of heat pumps, and it is physically interpretable. Due to the nonlinear characteristics of core components such as compressors and expansion valves, as well as the interference of ambient temperature, the identification accuracy of traditional linear models in heat pump systems is low. This invention integrates the modeling of nonlinear components such as compressors and expansion valves with ambient temperature correction to construct a heat pump Hammerstein-Wiener model, with compressor speed, expansion valve opening, and ambient temperature as inputs and hot water outlet temperature as output. To solve the problem of collaborative identification of high-dimensional parameters in strongly nonlinear systems, this invention adopts the Adam improved co-evolutionary algorithm, which explores the parameter subspace through multi-population division of labor and combines a cross-population elite fusion mechanism to significantly improve the search efficiency of high-dimensional parameter space while ensuring the integrity of identification.

[0068] (2) Due to the shortcomings of traditional co-evolutionary algorithms, such as population stagnation and local convergence, this invention introduces a mutation and Adam adaptive gradient optimization mechanism. The ergodicity of the Logistic map is used to generate the initial population, and chaotic mutation and Gaussian mutation are performed to ensure that the parameters cover the entire search space and avoid local optima. Furthermore, the Adam optimizer is introduced to replace traditional gradient descent, using first-order momentum to accelerate convergence and second-order momentum to adaptively adjust the step size, solving the problems of slow convergence and gradient oscillation in the later stages of the evolutionary algorithm. Figure 3 It can be seen that the improved co-evolutionary algorithm can better identify system model parameters.

[0069] (3) Adam's improved co-evolutionary algorithm can better identify the Hammerstein-Wiener model of heat pumps. Parameter estimation errors are significantly reduced, and the deviation between model simulation output and measured data is smaller. For example... Figure 3 As shown in the curve, this also fully demonstrates its good applicability to air source heat pump control systems. Attached Figure Description

[0070] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0071] Figure 1 This is a schematic diagram of the Hammerstein-Wiener model of the heat pump system in this invention.

[0072] Figure 2 This is a schematic diagram of a compressor model with saturation characteristics in this invention.

[0073] Figure 3The diagram illustrates the variation of the optimal fitness of the heat pump system before and after the algorithm improvement in this invention with the number of iterations T. CEA represents the algorithm identification curve before the improvement, and Adam-CEA represents the algorithm identification curve after the improvement. Detailed Implementation

[0074] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. Of course, the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0075] Example 1

[0076] This invention is achieved through the following measures: a method for identifying parameters of a heat pump system based on the Adam improved co-evolutionary algorithm, comprising the following steps:

[0077] See Figures 1 to 3 This embodiment provides a technical solution for heat pump system parameter identification based on the Adam improved co-evolutionary algorithm, and the specific steps are as follows:

[0078] Step 1) Construct a Hammerstein-Wiener model of the heat pump system based on the thermodynamic mechanism of its core components to obtain a physically interpretable identification model of the heat pump system.

[0079] Step 2) Constructing the identification process for the Adam improved co-evolutionary algorithm

[0080] The specific modeling steps for step 1) are as follows:

[0081] Step 1-1) Construct nonlinear models of the core components of the air source heat pump, including a nonlinear model of the compressor, a nonlinear model of the expansion valve, an ambient temperature correction model, and a thermodynamic mechanism model of the hot water outlet temperature:

[0082] Compressor nonlinear model: Based on thermodynamic mechanisms and the ideal gas law, a saturation function is introduced to characterize the physical saturation characteristics of the rotational speed. The model form is as follows:

[0083]

[0084] Where η v Let u1(t) be the volumetric efficiency, u1(t) be the compressor speed input, and V be the volumetric efficiency. th Theoretical displacement, ρ(t) is the refrigerant density. (ρ is pressure, m is molar mass, R is ideal gas constant, T is refrigerant temperature); according to Figure 2 A compressor model with saturation characteristics can be represented as:

[0085]

[0086] Where x1 and x2 are constants.

[0087] Nonlinear model of expansion valve: based on the pressure difference Δp = p across the valve. A -p B (p A p is the condenser outlet pressure. B Modeling the evaporator inlet pressure for both turbulent and laminar flow states, and introducing a pressure drop threshold. The model is integrated as follows:

[0088]

[0089] Where u2(t) is the expansion valve opening input, k val The expansion valve flow coefficient;

[0090] Ambient temperature correction model: An ambient temperature correction term is introduced to compensate for the decrease in the evaporator heat transfer coefficient. The equivalent input is:

[0091]

[0092] Where c1<0 is the ambient temperature correction factor, T a (t) represents the ambient temperature;

[0093] Hot water outlet temperature model: Combining the heat transfer and exponential saturation characteristics of the condenser, the model is as follows:

[0094]

[0095] Where T w，in ΔT represents the inlet temperature of the cold water. max The maximum temperature rise is given by k, which is a nonlinear coefficient, and T is the temperature rise. co This refers to the condenser temperature.

[0096] Steps 1-2) Integrate the compressor nonlinear model, expansion valve nonlinear model, ambient temperature correction model, and hot water outlet temperature thermodynamic mechanism model to construct a heat pump Hammerstein-Wiener model, based on... Figure 1 The system can be represented as:

[0097] x(t)=f1[u1(t)]+f2[u2(t)]+f3[u3(t)] (24)

[0098]

[0099] y(t)=h[w(t)] (26)

[0100] In the formula, t is the time variable, u(t) and y(t) are the system's input and output signals, respectively, and v(t) is the equivalent correction input for ambient temperature. x(t) is the output of the nonlinear part of the system, and w(t) is the linear dynamic part of the system. -1 It is the unit shift operator: q -1 =u(t-1). A(q) -1 B(q) -1 ) is q -1 The constant polynomial in the equation is defined as follows:

[0101]

[0102] Where a i It is a polynomial A(q) -1 The factor of ) b i It is a polynomial B(q) -1 The factor of ) and the order n are known. The relationship between the output w(t) and the output x(t) of the nonlinear part can be obtained from formula (6):

[0103]

[0104] Based on equations (1), (3), (4), and (6), they are updated as follows:

[0105]

[0106] According to equation (5), equation (8) is updated to:

[0107]

[0108] Where T w,in ΔT max k1, k2, and c1 are constants.

[0109] Steps 1-4) Define the parameter vector to be identified, integrating linear parameters, nonlinear coefficients, and physical parameters, in the form of:

[0110] θ = [a1, a2, ..., a n b1, b2, ..., b n [k1, k2, c1, Δp] T

[0111] Step 2) The specific process of the algorithm includes the following steps:

[0112] Step 2-1) Initialization settings: Given the number of iterations k, population size N, and data length M, define the parameter search space [lb]. j ub j (j is the parameter index); Three groups are divided into pole search groups (PS) θps = [a1, a2, ..., a n ] T Zero-point search group (ZS)θ zs = [b1, b2, ..., b n ] T Nonlinear parameter search group (NP)θ np =[k1,k2,c1,Δp] T The initial population is generated based on the Logistic chaotic mapping, and the chaotic sequence is mapped to the parameter space, as shown in the formula:

[0113] z n+1 =4·z n ·(1-z n ),z0∈(0,1),n=1,2,…,N (32)

[0114] Step 2-2) Data Acquisition: Use the compressor speed u1(t), expansion valve opening u2(t), and ambient temperature u3(t) of the air source heat pump as input data, and the hot water outlet temperature y(t) as output data, and record M sets of sample data;

[0115] Step 2-3) Fitness Function Definition: A stability penalty term is introduced to ensure system stability. The fitness function is:

[0116]

[0117] Where y meas (t) represents the measured output, y sim (θ,t) represents the model simulation output, r i For the poles of the CARMA model, n s For the number of extreme points; Steps 2-4) Explore the co-evolution of the three populations: subpopulations evolve independently, and the PS population fixes θ. zs and θ np , for θ ps Perform chaotic mutation (early stage) or Gaussian mutation (late stage); fix θ in the ZS group. ps and θ np , for θ zs Perform similar mutations; NP group fixes θ ps and θ zs , for θ np Perform similar mutations; the mutation formula is: Chaotic Mutation Gaussian mutation

[0118] Steps 2-5) Cross-population elite information fusion: Select the top M best individuals from each of the three populations to form an elite set. Replace the nonlinear parameters of the PS and ZS populations with the elites of the NP population. Replace the zero-point parameters of the PS and NP populations with the elites of the ZS population. Replace the extreme parameters of the ZS and NP populations with the elites of the PS population. After evaluation, retain the better solution.

[0119] Steps 2-6) Population Reconstruction and Global Optimal Update: Merge the three populations, select the top N best individuals by fitness, re-divide them into a new generation of three populations, and update the global optimal solution θ. * ;

[0120] Steps 2-7) Adam Optimization and Refinement: When the population shows no performance improvement for three consecutive generations, start the Adam optimizer to refine the globally optimal parameter θ. * Refining:

[0121] The gradient of the fitness function can be approximated by numerical differentiation:

[0122]

[0123] Where h is the differential step size, e j It is a unit vector.

[0124] Calculating the first-order momentum (momentum term) is used to accumulate historical gradient directions, accelerating convergence:

[0125]

[0126] Where β1 = 0.9 is the momentum decay coefficient, and t is the Adam iteration number.

[0127] Calculate second-order momentum (adaptive learning rate): Adjust the step size based on the squared gradient to suppress oscillations.

[0128]

[0129] It also compensates for bias, updates parameters using an adaptive learning rate, and refines the θ... * Inject the worst-performing individual into the population;

[0130] Step 2-8) Iteration Termination Judgment: If the maximum number of iterations K is reached or the fitness meets the preset accuracy requirement, output the identification result θ. * Complete system identification.

[0131] Example 2

[0132] Based on Example 1, the simplified Hammerstein-Wiener model of the heat pump system based on the thermodynamic mechanism of the core components used in this example is shown in the figure below. Figure 1As shown. Where u1(t), u2(t), and u3(t) are inputs, and y(t) is the output. Using the Hammerstein-Wiener model of the heat pump system mentioned above, the following model can be established for this embodiment:

[0133] Output stage:

[0134] y(t)=h(w(t))=T w,in +ΔT max ·(1-e -0.85w (t)) (19)

[0135] Where T w,in and ΔT max It is a constant. Comparing the above model with step 1), we can obtain...

[0136] k1 = -0.85

[0137] Intermediate linear static element:

[0138] Comparing the above model with step 1), we can obtain

[0139] a1=-0.13, a2=-0.2, a3=-0.50, b1=1.07, b2=0.83, b3=-0.41

[0140] Input process:

[0141]

[0142] Where k val It is a constant. Comparing the above model with step 1), we can obtain...

[0143] k2=0.12, c1=-0.015, Δp=600

[0144] For the above model, a fitness function is determined for use in this algorithm. The fitness function is defined as follows:

[0145]

[0146] Where y meas (t) represents the measured output, y sim (θ, t) represents the model simulation output, r i For the poles of the CARMA model, n s Let be the number of poles; to facilitate the substitution of the parameters to be identified into this algorithm, the parameters to be identified are grouped into a parameter vector θ, and the parameters to be identified are as follows:

[0147] θ=[a1, a2, a3, b1, b2, b3, k1, k2, c1, Δp] T

[0148] Based on the initialization in step 2-1), given the number of iterations k, the population size N, and the upper and lower bounds of the parameters, divide the population into three types.

[0149] Collect input and output data according to step 2-2);

[0150] Define the fitness function f(θ) according to steps 2-3);

[0151] Execute θ according to steps 2-4) ps ,θ zs ,θ np Exploring the co-evolution of three populations;

[0152] Based on the cross-population elite information fusion in steps 2-5), the better solution is evaluated and retained.

[0153] Based on steps 2-6), the three groups are merged and re-divided into a new generation of three groups, and the global optimal solution θ is updated. * ;

[0154] According to steps 2-7), when the population has no improvement for three consecutive generations, the Adam optimizer is started to optimize θ. * Refining;

[0155] Based on the iteration termination in steps 2-8), the identification result θ is output. * Complete system identification;

[0156] Several issues need to be considered when setting the upper and lower bounds of each parameter: Air source heat pump systems exhibit characteristics such as compressor saturation and expansion valve flow nonlinearity, while core physical parameters such as the expansion valve flow coefficient k... val The maximum temperature rise must conform to the actual range of engineering values; therefore, we will set specific value ranges for the relevant parameters. When selecting the population size, a trade-off between search accuracy and time consumption is necessary. When the population size is too small, the search time will decrease, but the parameter subspaces of the three populations will not be fully explored, potentially missing the optimal solution and affecting search accuracy. Conversely, when the population size is too large, the computational cost of cross-population elite fusion increases significantly, which, while enhancing global search capabilities, will substantially prolong the search time. Therefore, in actual parameter selection, we need to make appropriate adjustments based on the Hammerstein-Wiener model structure, the dimensions of the parameters to be identified, and the sample characteristics of the input and output data to achieve better performance.

[0157] A comparison of the identification performance of the heat pump system identification method based on the improved Adam co-evolutionary algorithm in this embodiment before and after the improvement is shown below. Figure 3As shown in the figure. The identification results demonstrate that, in terms of identification error and convergence speed, the method of this invention has the following advantages: as the number of iterations T increases, the identification error becomes smaller and eventually approaches zero. Therefore, this identification method can effectively identify the Hammerstein-Wiener model of the heat pump system, has a fast convergence speed, and easily escapes local optima. This also indicates that this identification method has good applicability to air source heat pump control systems.

[0158] Example 3: This example proposes a method for identifying parameters of a heat pump system based on the Adam improved co-evolutionary algorithm. The system includes:

[0159] The physically interpretable heat pump system identification model building module is configured to perform the following process: construct a physically interpretable heat pump system identification model based on the Hammerstein-Wiener model of the heat pump system with the thermodynamic mechanism of the core components;

[0160] The Adam improved co-evolutionary algorithm building module is configured to perform the following process: building the identification process of the Adam improved co-evolutionary algorithm.

[0161] Example 4: This example proposes an electronic system, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method steps of the present invention.

[0162] Example 5: This example proposes a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it implements the steps of the method described in this invention, which will not be repeated here.

[0163] Example 6: This example proposes a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, they implement the steps of the method described in this invention, which will not be repeated here.

[0164] It should be noted that the processing flow of embodiments 3-6 corresponds to the specific steps of the method provided in embodiment 1 of the present invention, and has the corresponding functional modules and beneficial effects of the method. Technical details not described in detail in this embodiment can be found in the method provided in embodiment 1 of the present invention.

[0165] The program code used to implement the methods of this application may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0166] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for parameter identification of a heat pump system based on an improved Adam co-evolutionary algorithm, characterized in that, Includes the following steps: Step 1) Construct a Hammerstein-Wiener model of a heat pump system based on the thermodynamic mechanism of its core components to obtain a physically interpretable identification model of the heat pump system; Step 2) Construct the identification process for the Adam improved co-evolutionary algorithm.

2. The method for identifying parameters of a heat pump system based on the Adam improved co-evolutionary algorithm according to claim 1, characterized in that, Step 1) specifically includes the following steps: Step 1-1) Construct nonlinear models of the core components of the air source heat pump, including a nonlinear model of the compressor, a nonlinear model of the expansion valve, an ambient temperature correction model, and a thermodynamic mechanism model of the hot water outlet temperature: Compressor nonlinear model: Based on thermodynamic mechanisms and the ideal gas law, a saturation function is introduced to characterize the physical saturation characteristics of the rotational speed. The model form is as follows: Where η v Let u1(t) be the volumetric efficiency, u1(t) be the compressor speed input, and V be the volumetric efficiency. th Theoretical displacement, ρ(t) is the refrigerant density. ρ is pressure, m is molar mass, R is ideal gas constant, and T is refrigerant temperature; the compressor model with saturation characteristics is represented as: Where x1 and x2 are constants; Nonlinear model of expansion valve: based on the pressure difference Δp = p across the valve. A -p B p A p is the condenser outlet pressure. B For the evaporator inlet pressure, model the flow under turbulent / laminar conditions and introduce a pressure drop threshold. The model is integrated as follows: Where u2(t) is the expansion valve opening input, k val The expansion valve flow coefficient; Ambient temperature correction model: An ambient temperature correction term is introduced to compensate for the decrease in the evaporator heat transfer coefficient. The equivalent input is: Where c1 < 0 is the ambient temperature correction factor, T a (t) represents the ambient temperature; Hot water outlet temperature model: Combining the heat transfer and exponential saturation characteristics of the condenser, the model is as follows: Where T w,in ΔT represents the inlet temperature of the cold water. max The maximum temperature rise is given by k, which is a nonlinear coefficient, and T is the temperature rise. co This refers to the condenser temperature. Steps 1-2) Integrate the compressor nonlinear model, expansion valve nonlinear model, ambient temperature correction model, and hot water outlet temperature thermodynamic mechanism model to construct a heat pump Hammerstein-Wiener model, which is represented as: x(t)=f1[u1(t)]+f2[u2(t)]+f3[u3(t)] (6); y(t)=h[w(t)] (8); In the formula, t is the time variable, u(t) and y(t) are the system's input and output signals, respectively, v(t) is the equivalent correction input for ambient temperature, x(t) is the output of the nonlinear part of the system, w(t) is the linear dynamic part of the system, and q -1 It is the unit shift operator: q -1 =u(t-1), A(q) -1 B(q) -1 ) is q -1 The constant polynomial in the equation is defined as follows: Where a i It is a polynomial A(q) -1 The factor of ) b i It is a polynomial B(q) -1 The factor of ) and the order n are known. The relationship between the output w(t) and the output x(t) of the nonlinear part is obtained from formula (6): Based on equations (1), (3), (4), and (6), they are updated as follows: According to equation (5), equation (8) is updated to: Where T w,in ΔT max k1, k2, and c1 are constants; Steps 1-3) Define the parameter vector to be identified, integrating linear parameters, nonlinear coefficients, and physical parameters, in the form of: θ=[a1,a2,...,a n ,b1,b2,...,b n ,k1,k2,c1,Δp] T 。 3. The heat pump control system identification method based on the Adam improved co-evolutionary algorithm according to claim 1, characterized in that, Step 2) specifically includes the following steps: Step 2-1) Initialization settings: Given the number of iterations k, population size N, and data length M, define the parameter search space [lb]. j ,ub j ], j is the parameter index; three groups are divided into pole search groups (PS) θ ps =[a1,a2,...,a n ] T Zero-point search group (ZS)θ zs =[b1,b2,…,b n ] T Nonlinear parameter search group (NP)θ np =[k1,k2,c1,Δp] T An initial population is generated based on the Logistic chaotic mapping, and the chaotic sequence is mapped to the parameter space, as shown in the formula: z n+1 =4·z n ·(1-z n ),z0∈(0,1),n=1,2,...,N (14); Step 2-2) Data Acquisition: Use the compressor speed u1(t), expansion valve opening u2(t), and ambient temperature u3(t) of the air source heat pump as input data, and the hot water outlet temperature y(t) as output data, and record M sets of sample data; Step 2-3) Fitness Function Definition: A stability penalty term is introduced to ensure system stability. The fitness function is: Where y meas (t) represents the measured output, y sim (θ,t) represents the model simulation output, r i For the poles of the CARMA model, n s The number of poles; Steps 2-4) Exploring the co-evolution of three populations: subpopulations evolve independently, and the PS population fixes θ. zs and θ np , for θ ps Perform chaotic mutation or Gaussian mutation; fix θ in the ZS group. ps and θ np , for θ zs Perform similar mutations; NP group fixes θ ps and θ zs , for θ np Perform similar mutations; the mutation formula is: Chaotic Mutation Gaussian mutation Steps 2-5) Cross-population elite information fusion: Select the top M best individuals from each of the three populations to form an elite set. Replace the nonlinear parameters of the PS and ZS populations with the elites of the NP population. Replace the zero-point parameters of the PS and NP populations with the elites of the ZS population. Replace the extreme parameters of the ZS and NP populations with the elites of the PS population. After evaluation, retain the better solution. Steps 2-6) Population Reconstruction and Global Optimal Update: Merge the three populations, select the top N best individuals by fitness, re-divide them into a new generation of three populations, and update the global optimal solution θ. * ; Steps 2-7) Adam Optimization and Refinement: When the population shows no performance improvement for three consecutive generations, start the Adam optimizer to refine the globally optimal parameter θ. * Refining: The gradient of the fitness function can be approximated by numerical differentiation: Where h is the differential step size, e j It is a unit vector; Calculating first-order momentum is used to accumulate historical gradient directions, accelerating convergence: Where β1 = 0.9 is the momentum decay coefficient, and t is the Adam iteration number; Calculate second-order momentum: Adjust the step size based on the squared gradient to suppress oscillations. It also compensates for bias, updates parameters using an adaptive learning rate, and refines the θ... * Inject the worst-performing individual into the population; Step 2-8) Iteration Termination Judgment: If the maximum number of iterations K is reached or the fitness meets the preset accuracy requirement, output the identification result θ. * Complete system identification.

4. A parameter identification system for a heat pump system based on the Adam improved co-evolutionary algorithm, characterized in that, The system includes: The physically interpretable heat pump system identification model building module is configured to perform the following process: construct a physically interpretable heat pump system identification model based on the Hammerstein-Wiener model of the heat pump system with the thermodynamic mechanism of the core components; The Adam improved co-evolutionary algorithm building module is configured to perform the following process: building the identification process of the Adam improved co-evolutionary algorithm.

5. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed, it implements the steps of the method as described in any one of claims 1 to 3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program is configured to implement the steps of the method according to any one of claims 1 to 3 when invoked by a processor.

7. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 3.

Images