Heat pump control system identification method based on improved SABO algorithm

Abstract

The invention provides a heat pump control system identification method based on an improved SABO algorithm, and belongs to the technical field of industrial control process system identification. The technical problem that an SABO algorithm is prone to falling into a local optimal solution in the system identification process under the condition of insufficient sample data and the system parameter identification process is solved, and the technical scheme includes the following steps that firstly, a single-input-single-output Hammerstein model of the air source heat pump control system with the dead zone characteristic is constructed, a Hammerstein system identification model of the air source heat pump control system with dead zone characteristics under the condition of insufficient sample data is obtained; and 2) constructing an identification process of the improved SABO algorithm. The air source heat pump control system parameter identification method has the advantages that the convergence speed is high, the identification precision is high, the output parameter estimation error is small, and meanwhile it is indicated that the identification method has good applicability to air source heat pump control system parameter identification.

Description

Technical Field

[0001] The invention relates to the technical field of heat pump control process system identification, and in particular to a heat pump control system identification method based on an improved SABO algorithm. Background Art

[0002] More than half of the energy consumption of buildings comes from HVAC systems. In order to reduce energy loss, it is particularly important to use energy-saving equipment with low cost and high efficiency. Heat pumps can effectively utilize low-grade thermal energy and provide a large amount of heating due to their ability to achieve reverse Carnot cycle under low power supply. Compared with ground source heat pumps and water source heat pumps, the heat device of air source heat pumps is in the air, with lower installation and operation costs and lower energy consumption. In order to better study the air source heat pump control system, it is necessary to establish a system model for the air source heat pump control system and identify the parameters of the established model.

[0003] The air source heat pump control system exhibits nonlinear characteristics. The block structure nonlinear system has excellent research characteristics and can be used to describe most nonlinear dynamic systems. The Wiener-Hammerstein (WH) system is an important branch of the block structure nonlinear system. It consists of two parts: static nonlinearity and delayed dynamic linearity. It can be used to describe many actual nonlinear systems. The identification principle of the WH system is to establish a mathematical model of each series link of the system, and use the optimization method to solve the parameters of each link based on the measured input and output data. Common identification methods include iterative identification method, least squares parameter identification method and frequency domain method. The paper Enhanced Identification Techniques for Wiener-Hammerstein Models in Dynamic Systems proposes a complex system identification method for optimizing the WH model; the paper Enhanced Wiener-Hammerstein Model Identification for Nonlinear Systems proposes an enhanced WH model identification method. Compared with the above literature, most of the WH methods discussed have complete input and output data, ignoring the situation of insufficient data in actual control systems.

[0004] The Subtraction Mean Optimizer (SABO) algorithm uses the subtraction mean of multiple agents to update the positions of population members in the search space, such as the average value, the difference in the positions of the search agents, and the difference in the signs of the two values ​​of the objective function. It uses the arithmetic mean position of all search agents to group the branch variables in the mixed integer programming and process them simultaneously. This method expands the scope of exploration, reduces the number of iterations, and speeds up the solution process. However, when the processing task is complex, the computational complexity of the SABO algorithm increases with the increase in the number of iterations, and it is easy to fall into the local optimal solution.

[0005] How to solve the above technical problems is the subject faced by the present invention. Summary of the invention

[0006] The purpose of the present invention is to provide a heat pump control system identification method based on an improved SABO algorithm, which has an efficient and optimal search capability, so the identification speed is fast, and effectively overcomes the shortcomings of the traditional SABO algorithm that the computational complexity increases when the number of iterations increases and it is easy to fall into the local optimal solution.

[0007] The present invention is achieved by the following measures: A heat pump control system identification method based on an improved SABO algorithm (SABO-CR), comprising the following steps:

[0008] Step 1) construct a single-input single-output Hammerstein model of an air source heat pump control system with a dead zone characteristic, and obtain a Hammerstein system identification model of the air source heat pump control system with a dead zone characteristic under the condition of insufficient sample data;

[0009] Step 2) Construct the identification process of the improved SABO algorithm.

[0010] As a further optimization scheme of the heat pump control system identification method based on the improved SABO algorithm provided by the present invention, the step 1) specifically includes the following steps:

[0011] Step 1-1) Construct a single-input single-output Hammerstein model for the air source heat pump control system: According to Figure 1 , the Hammerstein model can be expressed as:

[0012] x(t)=h[u(t)] (1)

[0013]

[0014] w(t)=D(q -1 )v(t) (3)

[0015] Where t is a time variable, u(t) and y(t) are the system input and output signals, respectively, x(t) is the output of the nonlinear part of the system, w(t) is an unmeasurable intermediate signal, and v(t) represents a random Gaussian white noise with zero mean. -1 is the unit shift operator: q -1 u(t)=u(t-1), A(q -1 ), B(q -1 )、D(q -1 ) is a constant polynomial in q with the following definition:

[0016]

[0017] Among them, a i is the polynomial A(q -1 ), b i is the polynomial B(q -1 ), d i is the polynomial D(q -1 ) factors, the order n is known;

[0018] Step 1-2) From equation (2), we can get the relationship between the output y(t) and the input u(t) and the output x(t) of the nonlinear part:

[0019] A(q -1 )y(t)=B(q -1 )x(t)+A(q -1 )w(t) (7)

[0020] Formula (2) can be reformulated as:

[0021]

[0022] Steps 1-3) According to Figure 2 , the nonlinear part of the air source heat pump control system can be expressed as:

[0023]

[0024] Among them, t1, t2, and x1 are all constants. c=-kt1, introduce a switch function:

[0025]

[0026] Formula (9) can be reformulated as:

[0027]

[0028] As a further optimization scheme of the heat pump control system identification method based on the improved SABO algorithm provided by the present invention, the model of step 1) is a single-input single-output Hammerstein model of an air source heat pump control system with a dead zone characteristic.

[0029] As a further optimization scheme of the heat pump control system identification method based on the improved SABO algorithm provided by the present invention, the specific steps of step 2) constructing the identification process of the improved SABO algorithm are as follows:

[0030] Step 2-1) Initialize parameters, define population size, and set upper and lower bounds of n parameters;

[0031] Step 2-2) using the regulating valve opening of the air source heat pump control system as input data of the heat pump system, and using the hot water temperature of the air source heat pump control system as output data of the heat pump system;

[0032] Step 2-3) Define the fitness function fitness(L):

[0033] fitness=[Y(t)-y(t)] 2 (12)

[0034] Among them, Y(t) is the actual output value of the system, and y(t) is the estimated value of the system output;

[0035] Step 2-4) Pass in the input data and output data in the actual small sample sampling data, set the calculation time step time_steps, and initialize the theoretical model variables y(t) and x(t) to store the theoretical values:

[0036] y(t)=zeros(1,time_steps) (13)

[0037] x(t)=zeros(1,time_steps) (14)

[0038] Among them, y(t) represents the theoretical hot water temperature of the air source heat pump control system, x(t) is the nonlinear part of the air source heat pump control system, and zeros(1, time_steps) represents the generation of a zero matrix of 1*time_steps;

[0039] Step 2-5) Generate random noise v(t) with a mean of 0 and a standard deviation of 1, and use formula (12) to calculate the sum of squares of the errors between the actual output value and the theoretical output value as the fitness value, which is used to improve the fitness optimization iteration in the SABO algorithm;

[0040] Step 2-6) Define the SABO function, initialize the population, randomly generate individuals X between the upper and lower bounds of the parameters, initialize the fitness fit, and calculate the fitness value of each individual;

[0041] Step 2-7) Introduce the differential variation factor, initialize the optimal fitness, differential variation factor F and crossover rate CR;

[0042] Step 2-8) Perform differential mutation on each random individual X. The present invention selects three individuals X1, X2, and X3 for mutation operation:

[0043] V=X1+F·(X2-X3) (15)

[0044] Among them, V is the mutation vector;

[0045] Step 2-9) Traverse the n parameters of the current individual and perform mutation operations. Compare the random number rand with the crossover rate. If rand < CR, replace the corresponding parameter in V to obtain a new individual U.

[0046] Step 2-10) Calculate the fitness of the new individual and update the individual with better fitness. Save the current optimal solution fbest and the corresponding individual xbest, perform iterative optimization, output the optimal solution fbest as the fitness value, and output the optimal individual xbest as the values of the n parameters to complete the system parameter identification.

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0048] (1) The present invention establishes a parameter identification model for an air-source heat pump control system under insufficient sample data. The friction of the regulating valve causes the heat pump control system to have a dead zone nonlinear characteristic. By introducing a switching function, a single-input single-output Hammerstein model of the air-source heat pump control system with dead zone characteristics is derived, and the opening of the regulating valve is used as the input data of the air-source heat pump control system. In order to deal with the phenomenon that it is difficult to identify system parameters caused by insufficient sample data, the present invention uses an improved SABO algorithm, which can iteratively optimize to obtain the optimal system parameters according to scarce data and complete the identification of the system. Because of its extremely high calculation speed, the calculation efficiency of the identification method is greatly improved.

[0049] (2) The present invention introduces a differential mutation factor to enhance the global search ability of the SABO algorithm, help it jump out of the local optimal solution, and further improve the optimization efficiency. As Figure 3 can be seen, the improved SABO algorithm can better identify the system model parameters.

[0050] (3) The improved SABO algorithm of the present invention can better identify the single-input single-output nonlinear system of the heat pump, with high identification accuracy and small parameter estimation error, only 0.0696. As shown by the SABO-CR curve in Figure 3 , this also shows that this identification method has good applicability to the air-source heat pump control system. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention.

[0052] Figure 1 It is a schematic diagram of a single-input single-output Hammerstein general model for identifying an air-source heat pump control system by the improved SABO algorithm in the present invention.

[0053] Figure 2It is a schematic diagram of the nonlinear part of the air source heat pump control system with dead zone characteristics in the present invention.

[0054] Figure 3 It is a schematic diagram of the curve of the optimal fitness (identification error) of the air source heat pump control system identified with the number of iterations T before and after the algorithm improvement in the present invention; wherein, SABO represents the algorithm identification curve before improvement, and SABO-CR represents the algorithm identification curve after improvement. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Of course, the specific embodiments described here are only used to explain the present invention and are not used to limit the present invention.

[0056] Example 1

[0057] The technical solution provided in this embodiment is a method for identifying an air source heat pump control system based on an improved SABO algorithm, and the specific steps are as follows:

[0058] Step 1) construct a single-input single-output Hammerstein model of an air source heat pump control system with a dead zone characteristic, and obtain a Hammerstein system identification model of the air source heat pump control system with a dead zone characteristic under the condition of insufficient sample data;

[0059] Step 2) Construct the identification process of the improved SABO algorithm.

[0060] Specifically, the step 1) comprises the following steps:

[0061] Step 1-1) Construct a single-input single-output Hammerstein model for the air source heat pump control system: According to Figure 1 , the Hammerstein model can be expressed as:

[0062] x(t)=h[u(t)] (1)

[0063]

[0064] w(t)=D(q -1 )v(t) (3)

[0065] Where t is a time variable, u(t) and y(t) are the system input and output signals, respectively, x(t) is the output of the nonlinear part of the system, w(t) is an unmeasurable intermediate signal, and v(t) represents a random Gaussian white noise with zero mean. -1 is the unit shift operator: q -1 u(t)=u(t-1), A(q-1 ), B(q -1 )、D(q -1 ) is a constant polynomial in q with the following definition:

[0066]

[0067] Among them, a i is the polynomial A(q -1 ), b i is the polynomial B(q -1 ), d i is the polynomial D(q -1 ) factors, the order n is known;

[0068] Step 1-2) From equation (2), we can get the relationship between the output y(t) and the input u(t) and the output x(t) of the nonlinear part:

[0069] A(q -1 )y(t)=B(q -1 )x(t)+A(q -1 )w(t) (7)

[0070] Formula (2) can be reformulated as:

[0071]

[0072] Steps 1-3) According to Figure 2 , the nonlinear part of the air source heat pump control system can be expressed as:

[0073]

[0074] Among them, t1, t2, and x1 are all constants. c=-kt1, introduce a switch function:

[0075]

[0076] Formula (9) can be reformulated as:

[0077]

[0078] Specifically, the model of step 1) is a single-input single-output Hammerstein model of an air source heat pump control system with a dead zone characteristic.

[0079] Specifically, the specific steps of step 2) constructing the identification process of the improved SABO algorithm are as follows:

[0080] Step 2-1) Initialize parameters, define population size, and set upper and lower bounds of n parameters;

[0081] Step 2-2) using the regulating valve opening of the air source heat pump control system as input data of the heat pump system, and using the hot water temperature of the air source heat pump control system as output data of the heat pump system;

[0082] Step 2-3) Define the fitness function fitness(L):

[0083] fitness=[Y(t)-y(t)] 2 (12)

[0084] Among them, Y(t) is the actual output value of the system, and y(t) is the estimated value of the system output;

[0085] Step 2-4) Pass in the input data and output data in the actual small sample sampling data, set the calculation time step time_steps, and initialize the theoretical model variables y(t) and x(t) to store the theoretical values:

[0086] y(t)=zeros(1,time_steps) (13)

[0087] x(t)=zeros(1,time_steps) (14)

[0088] Among them, y(t) represents the theoretical hot water temperature of the air source heat pump control system, x(t) is the nonlinear part of the air source heat pump control system, and zeros(1, time_steps) represents the generation of a zero matrix of 1*time_steps;

[0089] Step 2-5) Generate random noise v(t) with a mean of 0 and a standard deviation of 1, and use formula (12) to calculate the sum of squares of the errors between the actual output value and the theoretical output value as the fitness value, which is used to improve the fitness optimization iteration in the SABO algorithm;

[0090] Step 2-6) Define the SABO function, initialize the population, randomly generate individuals X between the upper and lower bounds of the parameters, initialize the fitness fit, and calculate the fitness value of each individual;

[0091] Step 2-7) Introduce the differential variation factor, initialize the optimal fitness, differential variation factor F and crossover rate CR;

[0092] Step 2-8) Perform differential mutation on each random individual X. The present invention selects three individuals X1, X2, and X3 for mutation operation:

[0093] V=X1+F·(X2-X3) (15)

[0094] Among them, V is the mutation vector;

[0095] Step 2-9) Traverse the n parameters of the current individual and perform mutation operations. Compare the random number rand with the crossover rate. If rand < CR, replace the corresponding parameter in V to obtain a new individual U.

[0096] Step 2-10) Calculate the fitness of the new individual and update the individual with better fitness. Save the current optimal solution fbest and the corresponding individual xbest, perform iterative optimization, output the optimal solution fbest as the fitness value, and output the optimal individual xbest as the values of the n parameters to complete the system parameter identification.

[0097] Example 2

[0098] Based on Example 1, the schematic diagram of the air source heat pump control system with dead zone characteristics adopted in this example is as Figure 1 shown. Among them, u(t) is the opening of the regulating valve, x(t) is the state variable of the nonlinear part of the system, and y(t) is the hot water temperature.

[0099] Through the general single-input single-output Hammerstein model mentioned above, the following model can be established for this example:

[0100] A(q -1 )y(t) = B(q -1 )x(t) + w(t)A(q -1 ) (16)

[0101] A(q -1 ) = 1 + a1q -1 + a2q -2 + a3q -3 = 1 - 1.32q -1 + 0.52q -2 - 0.19q -3 (17)

[0102] B(q -1 ) = b1q -1 + b2q -2 + b3q -3 = 0.49q -1 - 0.33q -2 + 0.37q -3 (18)

[0103] D(q -1 ) = 1 + d1q -1 + d2q -2 + d3q -3 = 1 - 0.90q -1 + 1.45q -2 + q -3 (19)

[0104] Comparing the above model with step 1), we can get

[0105] a1=-1.32, a2=0.52, a3=-0.19,

[0106] b1=0.49, b2=-0.33, b3=0.37,

[0107] d1=-0.90, d2=1.45, d3=1.00,

[0108] For the above model, a fitness function fitness(L) is determined for use in the improved SABO algorithm. The fitness function is defined as follows:

[0109] fitness(L)=[Y(t)-y(t)] 2 (20)

[0110] Among them, Y(t) is the actual output value of the system, and y(t) is the estimated value of the system output.

[0111] In order to facilitate the introduction of the parameters to be identified into the improved SABO algorithm, the parameters to be identified are combined into a parameter vector L, and the parameters to be identified are as follows:

[0112] L=[t1,t2,x1,a1,a2,a3,b1,b2,b3,d1,d2,d3]∈R (21)

[0113] Initialize the population size according to step 2-1) and set the upper and lower bounds of the parameters;

[0114] Collect input and output data according to step 2-2);

[0115] Define the fitness function fitness(L) according to step 2-3);

[0116] According to step 2-4), input data and output data in the actual small sample sampling data are passed in;

[0117] According to step 2-5), introduce zero-mean random Gaussian white noise as the noise interference term of the system;

[0118] According to step 2-6), the sum of square errors between the actual output value and the theoretical output value is calculated as the fitness value;

[0119] Define the SABO function according to step 2-7) and randomly generate individual X;

[0120] According to step 2-8), introduce the differential mutation factor, set the differential mutation factor F and the crossover rate CR, and perform the mutation operation;

[0121] According to step 2-9), record the minimum fitness value fit and the corresponding optimal individual xbest;

[0122] According to step 2-10), the loop is completed, and the parameter estimation vector L obtained by identification is output, and the identification is completed.

[0123] Among them, several issues need to be considered when setting the upper and lower bounds of each parameter: Due to the partial nonlinear characteristics of the regulating valve in the air source heat pump control system, there is a certain relationship between the current output hot water temperature and the hot water temperature of the previous second, and the hot water temperature shows an upward trend in the working area of ​​the heat pump control system. The correlation coefficient needs to be no less than 1. Therefore, we will set a specific value range for the relevant parameters. When selecting the population size, it is necessary to weigh the search accuracy and time consumption. When the population size is too small, the search time will be reduced, but it may affect the search accuracy. Conversely, when the population size is too large, the search capability is enhanced, but the search time will increase. Therefore, when actually selecting parameter values, we need to make appropriate adjustments based on the characteristics of the model structure and input and output data to achieve better performance.

[0124] Comparison of algorithm identification effects before and after improvement of the heat pump control system identification method based on the improved SABO algorithm of this embodiment. Figure 3 As shown in Figure 2, the parameter identification results of the improved SABO algorithm are as follows: Figure 3 The SABO-CR curve is shown. The identification results show that the method of the present invention has the following advantages in terms of identification error and convergence speed: as the number of iterations T increases, the identification error becomes smaller and smaller and eventually tends to zero. Therefore, the identification method can effectively identify the Hammerstein system of the heat pump control system, has a fast convergence speed, and is easy to jump out of the local optimal solution. This also shows that the identification method has good applicability for air source heat pump control systems.

[0125] 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 principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A heat pump control system identification method based on an improved SABO algorithm, characterized in that: It includes the following steps: Step 1) Construct a single-input single-output Hammerstein model of an air-source heat pump control system with dead zone characteristics, and obtain a Hammerstein system identification model of an air-source heat pump control system with dead zone characteristics under the condition of insufficient sample data; Step 2) Construct the identification process of the improved SABO algorithm.

2. The heat pump control system identification method based on the improved SABO algorithm according to claim 1 is characterized in that: The said Step 1) includes the following steps: Step 1-1) Construct a single-input single-output Hammerstein model of an air-source heat pump control system, and the Hammerstein model is expressed as: x(t) = h[u(t)] (1) w(t)=D(q -1 )v(t) (3) Where t is the time variable, u(t) and y(t) are the system input and output signals, respectively, x(t) is the output of the nonlinear part of the system, w(t) is an unmeasurable intermediate signal, v(t) represents a random Gaussian white noise with zero mean, and q -1 is the unit shift operator: q -1 u(t)=u(t-1), A(q -1 ), B(q -1 )、D(q -1 ) is a constant polynomial in q, defined as follows: Among them, a i is the polynomial A(q -1 ), b i is the polynomial B(q -1 ), d i is the polynomial D(q -1 ) factors, the order n is known; Step 1-2) Obtain the relationship between the output y(t) and the input u(t) and the output x(t) of the nonlinear part from Equation (2): A(q -1 )y(t)=B(q -1 )x(t)+A(q -1 )w(t) (7) Equation (2) is re-expressed as: The nonlinear part of the air-source heat pump control system is expressed as: Among them, t1, t2, and x1 are all constants. c=-kt1, introduce a switch function: Equation (9) is re-expressed as:

3. The heat pump control system identification method based on the improved SABO algorithm according to claim 1 is characterized in that: The said Step 2) includes the following steps: Step 2-1) Initialize the parameters, define the population size, and set the upper and lower bounds of n parameters; Step 2-2) Take the opening degree of the regulating valve of the air-source heat pump control system as the input data of the heat pump system, and take the hot water temperature of the air-source heat pump control system as the output data of the heat pump system; Step 2-3) Define the fitness function fitness(L): fitness=[Y(t)-y(t)] 2 (12) where, Y(t) is the actual output value of the system, and y(t) is the estimated value of the system output; Step 2-4) Input the input data and output data in the actual small sample sampling data, set the calculation time step time_steps, and initialize the theoretical model variables y(t) and x(t) for storing the theoretical values: y(t) = zeros(1,time_steps) (13) x(t) = zeros(1,time_steps) (14) where, y(t) represents the theoretical hot water temperature of the air-source heat pump control system, x(t) is the nonlinear part of the air-source heat pump control system, and zeros(1,time_steps) represents generating a 1*time_steps zero matrix; Step 2-5) Generate random noise v(t) with a mean of 0 and a standard deviation of 1, and use Equation (12) to calculate the sum of squared errors between the actual output value and the theoretical output value as the fitness value for fitness optimization iteration in the improved SABO algorithm; Step 2-6) Define the SABO function, initialize the population, randomly generate an individual X between the upper and lower bounds of the parameters, and initialize the fitness fit for calculating the fitness value of each individual; Step 2-7) Introduce the differential mutation factor, and initialize the optimal fitness and the differential mutation factors F and the crossover rate CR; Step 2-8) Perform differential mutation on each random individual X, and select three individuals X1, X2, and X3 for mutation operation: V = X1 + F·(X2 - X3) (15) where, V is the mutation vector; Step 2-9) Traverse the n parameters of the current individual for mutation operation, compare with the crossover rate using the random number rand, and if rand < CR, replace the corresponding parameter in V to obtain a new individual U; Step 2-10) Calculate the fitness of the new individual and update the individuals with better fitness, save the current optimal solution fbest and the corresponding individual xbest, iteratively optimize, output the optimal solution fbest as the fitness value, output the optimal individual xbest as the value of n parameters, and complete the system parameter identification.

Images