A building load prediction method based on multi-connected air conditioning system operation data

Abstract

This invention discloses a building load prediction method based on the operating data of a multi-split air conditioning system, relating to the field of building load prediction. The method includes the following steps: establishing a multi-split energy efficiency data calculation model and obtaining multi-split energy efficiency data; establishing a first-order thermal dynamic gray box model of the building; constructing a particle swarm optimization (PSO) fitness function based on the accuracy evaluation index of the first-order thermal dynamic gray box model; solving for the model parameters using the PSO algorithm; establishing a second-order thermal dynamic gray box model of the building and setting the model parameter range; solving for the second-order model parameters using the PSO algorithm; and calculating the building load based on the solution results of the second-order model parameters. This invention, based on the capacity data of direct expansion air conditioning systems, can obtain stable building thermal dynamic model parameter identification results. It is applicable to the establishment of thermal dynamic models for any building equipped with a direct expansion air conditioning system and solves the problem of inaccurate automated parameter identification for the fluctuating capacity of direct expansion air conditioning systems.

Description

Technical Field

[0001] This invention relates to the field of building load forecasting, and more particularly to a building load forecasting method based on the operating data of a multi-split air conditioning system. Background Technology

[0002] The development of building big data and artificial intelligence technologies in recent years has provided a data foundation and modeling analysis algorithms for predicting the performance of air conditioning systems. Data-driven approaches to building dynamic building models offer advantages such as fast modeling speed and high prediction accuracy.

[0003] Gray box models possess both physical and data-driven characteristics. The physical characteristics stem from the fact that their input parameters need to be calculated within a given physical process model to obtain the building's thermal dynamics. The data characteristics arise from the fact that the descriptive parameters of their physical processes are constructed based on a large amount of existing operational data. Typically, the construction of gray box models relies on air conditioning operating data, particularly information such as cooling / heating power and temperature.

[0004] In multi-split air conditioning systems, the operating patterns of system equipment are characterized by large fluctuations and rapid changes, and the coordination states of the system equipment are highly variable and complex. Modeling using high-resolution data of the cooling / heating power of a multi-split system is prone to getting trapped in local optima, leading to model identification failure. Existing techniques cannot utilize the high-frequency, wide-range oscillation output values ​​of direct expansion air conditioners for stable identification of building thermal dynamic parameters, which can lead to local optima, rendering the identification model unusable during normal operations or causing it to fail during online updates.

[0005] Heuristic algorithms, such as the Particle Swarm Optimization (PSO) algorithm, have significant advantages over traditional mathematical programming methods when searching for solutions within non-convex feasible regions. In online identification processes, neural networks can periodically update building thermal dynamics models to better suit real-world needs, and the PSO algorithm offers greater adaptability to the entire identification process.

[0006] Therefore, those skilled in the art are dedicated to developing a building load forecasting method based on the operating data of multi-split air conditioning systems. Summary of the Invention

[0007] In view of the above-mentioned deficiencies of the prior art, the technical problem to be solved by the present invention is that the output value of the high-frequency wide-range oscillation capability of direct expansion air conditioner cannot be used to stably identify building thermal dynamic parameters, and the identification model cannot be used in normal processes.

[0008] To achieve the above objectives, the present invention provides a building load forecasting method based on multi-split air conditioning system operation data, characterized in that the method includes the following steps:

[0009] S101: Establish a calculation model for the energy efficiency data of multi-split air conditioners and obtain energy efficiency data of multi-split air conditioners;

[0010] S103: Establish a first-order thermal dynamic gray box model of the building;

[0011] S105: Construct the particle swarm optimization algorithm fitness function based on the accuracy evaluation index of the first-order thermal dynamic gray box model of the building.

[0012] S107: Using the particle swarm optimization algorithm, solve for the parameters of the first-order thermal dynamic gray box model of the building;

[0013] S109: Establish a second-order thermal dynamic gray box model of the building and set the parameter range of the second-order thermal dynamic gray box model of the building;

[0014] S111: Using the particle swarm optimization algorithm, solve for the parameters of the second-order thermal dynamic gray box model of the building;

[0015] S113: Based on the solution results of the second-order thermal dynamic gray box model parameters of the building, calculate the building load. Further, in step S101, the multi-split unit energy efficiency data calculation model is:

[0016] Q VRF =F idu (h in -h out ) / 1000

[0017] F idu It is calculated using the following formula:

[0018]

[0019] Among them, Q VRF To enhance the cooling capacity of the multi-split indoor unit, F idu For the refrigerant flow rate entering the multi-split indoor unit, h in h represents the enthalpy of the refrigerant entering the indoor unit. out F represents the enthalpy of the refrigerant flowing out of the indoor unit. comp F is the refrigerant mass flow rate at the compressor outlet. bypass,i Let i be the refrigerant mass flow rate of the bypass loop, and n be the number of bypass loops.

[0020] Furthermore, step S101 also includes the collection and preprocessing of the operating data of the multi-split air conditioning system. The preprocessing includes data cleaning and data completion.

[0021] The data cleaning is set to remove abnormal data, which includes negative sensor data and abnormal transmission data. The abnormal transmission data is filtered based on the rated capacity of the multi-split air conditioner.

[0022] The data completion uses a linear interpolation method to complete a small number of broken data points in the collected data. The linear interpolation method is as follows:

[0023]

[0024] Where N is the number of missing points, Y n For the data value of the nth missing point, Y0 and Y... N+1 These are the data values ​​immediately before and after the missing sequence.

[0025] Furthermore, the data acquisition period for the multi-unit refrigeration capacity data is 30 seconds, and the upper limit of the filtering threshold for the multi-unit refrigeration capacity value is 10. 6 The upper limit of the screening threshold for the temperature value is 50℃.

[0026] Furthermore, the first-order thermal dynamic gray box model of the building is as follows:

[0027]

[0028] Where C is the building's total lumped heat capacity, is the C-parameter of the building's first-order thermal dynamic gray box model, and R... opaque and R window Let T be the thermal resistance of the building envelope, R be the parameter of the first-order thermal dynamic gray box model of the building, and T be the thermal resistance of the building envelope. in Indoor temperature, T out For the outdoor temperature, the optimization range for parameters C and R is set to positive real numbers only, without setting an upper limit.

[0029] Further, in step S105, the accuracy evaluation index of the first-order thermal dynamic gray box model of the building takes the minimum absolute value of the difference between the predicted indoor temperature and the actual indoor temperature as the objective function, and the objective function is:

[0030]

[0031] Where n is the sample number and N is the sample size. This represents the actual indoor temperature. This represents the predicted indoor temperature.

[0032] Further, in step S107, the particle swarm algorithm includes the following steps:

[0033] S1071: Set the built-in parameters of the particle swarm optimization algorithm, including particle population size, maximum number of iterations, learning factor and external solution set size;

[0034] S1072: Randomly initialize the velocity and position of the particle swarm;

[0035] S1073: Substitute the particles into the first-order thermal dynamic gray box model of the building, calculate the fitness function of each particle based on the fitness function of the particle swarm algorithm, and obtain the fitness function value of the particle.

[0036] S1074: Determine the relationship between the current fitness value of each particle and its individual best value pbest. If the fitness value of this generation of particles is superior, do not update it; otherwise, update the individual best value pbest. Compare pbest with the global best value gbest. If pbest is superior, update the gbest value.

[0037] S1075: Update the next generation of particles according to the particle renewal rate formula;

[0038] S1076: Repeat steps S1073 to S1075 until the set minimum error is met or the maximum number of iterations is reached, to obtain the parameter values ​​of the first-order thermal dynamic gray box model of the building.

[0039] Furthermore, the particle update rate formula in the particle swarm optimization algorithm is:

[0040]

[0041]

[0042] in, It is the d-th dimension component of the velocity vector of particle i in the k-th iteration. pbest is the d-th dimension component of the position of particle i in the k-th iteration. id It is the d-th dimension component of the historical individual optimal value position of particle i, gbest id ω is the d-th dimension component of the historical global optimal position of all particles, c1 and c2 are acceleration constants, and r1 and r2 are random functions.

[0043] Further, in step S109, the second-order thermal dynamic gray box model of the building is:

[0044]

[0045]

[0046] Based on the R and C parameters in the first-order thermal dynamic gray box model of the building, the optimization range of the R and C parameters in the second-order thermal dynamic gray box model of the building is set as follows:

[0047]

[0048] C air <10 log(C)+1 C wall >10log(C)+1

[0049] Wherein, parameter C air and C wall For parameter C in the second-order thermal dynamic gray box model of the building, C air For the building air heat capacity, C wall For the total heat capacity of the building envelope, parameters R1 and R2 are given. wall，outer and R wall，inner R is the parameter in the second-order thermal dynamic gray box model of the building, where R1 is the thermal resistance term directly related to the indoor and outdoor dry-bulb temperatures. wall，inner R represents the overall thermal resistance of the building's integrated building envelope. wall，outer For the overall thermal resistance of the building envelope on the outside, Tw all Q represents the temperature of the building's lumped envelope. other For other loads.

[0050] Further, in step S113, the building load is calculated in the following manner:

[0051]

[0052] Among them, Q building For building load.

[0053] In a preferred embodiment of the present invention, the present invention has the following advantages over the prior art:

[0054] 1. The two-stage building thermal dynamic gray box model identification method proposed in this invention enables stable building thermal dynamic model parameter identification results to be obtained based on direct expansion air conditioning capacity data, and maintains good identification stability.

[0055] 2. This invention can be universally applied to the establishment of thermal dynamic models for any building equipped with a direct expansion air conditioning system, and solves the problem that the fluctuation capacity of the direct expansion air conditioning system cannot be accurately identified by automated parameters.

[0056] The following will further explain the concept, specific structure, and technical effects of the present invention in conjunction with the accompanying drawings, so as to fully understand the purpose, features, and effects of the present invention. Attached Figure Description

[0057] Figure 1 This is a flowchart of a preferred embodiment of the building load prediction method of the present invention;

[0058] Figure 2 This is a flowchart of the construction process of a second-order thermal dynamic gray box model of a building according to a preferred embodiment of the present invention;

[0059] Figure 3This is a schematic diagram of the refrigerant piping of a multi-split air conditioning system according to a preferred embodiment of the present invention;

[0060] Figure 4 This is a schematic diagram of the refrigeration capacity data of a multi-unit air conditioning system according to a preferred embodiment of the present invention;

[0061] Figure 5 This is a comparison of dynamic simulation results based on the capacity data of a direct expansion air conditioner according to a preferred embodiment of the present invention. Detailed Implementation

[0062] The following description, with reference to the accompanying drawings, illustrates several preferred embodiments of the present invention to make its technical content clearer and easier to understand. The present invention can be embodied in many different forms, and the scope of protection of the present invention is not limited to the embodiments mentioned herein.

[0063] In the accompanying drawings, components with the same structure are indicated by the same numerical designation, and components with similar structures or functions are indicated by similar numerical designations. The dimensions and thicknesses of each component shown in the drawings are arbitrary, and the present invention does not limit the dimensions and thicknesses of each component. To make the illustrations clearer, the thickness of some components has been appropriately exaggerated in the drawings.

[0064] like Figure 1 , Figure 2 As shown in the figure, the present invention provides a building load prediction method based on the operation data of a multi-split air conditioning system. This method establishes a second-order gray-box mathematical model of the building and constrains the optimization range of the second-order model parameters using the parameter identification results of the first-order gray-box model. Optimization is then performed using a particle swarm optimization algorithm. This method can provide a method for constructing a thermal dynamic model of a building under the operation data of a multi-split air conditioning system, improving the feasibility and accuracy of building load prediction. The building load prediction method includes the following steps:

[0065] S101: Establish a multi-split air conditioner energy efficiency data calculation model and obtain multi-split air conditioner energy efficiency data.

[0066] The energy efficiency data calculation model for this multi-split air conditioner is as follows:

[0067] Q VRF =F idu (h in -h out ) / 1000

[0068] F idu It is calculated using the following formula:

[0069]

[0070] Among them, Q VRF To enhance the cooling capacity of the multi-split indoor unit, F iduFor the refrigerant flow rate entering the multi-split indoor unit, h in h represents the enthalpy of the refrigerant entering the indoor unit. out F represents the enthalpy of the refrigerant flowing out of the indoor unit. comp F is the refrigerant mass flow rate at the compressor outlet. bypass,i Let i be the refrigerant mass flow rate of the bypass loop, and n be the number of bypass loops.

[0071] To obtain multi-split air conditioning system energy efficiency data, the above steps also include the collection and preprocessing of operating data from the multi-split air conditioning system. The data collection cycle for the cooling capacity of the multi-split system is 30 seconds. Preprocessing includes data cleaning and data completion.

[0072] Data cleaning is set to remove outlier data, including negative sensor values ​​and transmission anomalies. The screening threshold for transmission anomalies is set based on the rated capacity of the multi-split air conditioner, with an upper limit of 10 for the screening threshold of the multi-split air conditioner's cooling capacity. 6 The upper limit of the temperature value screening threshold is 50℃;

[0073] Data completion uses linear interpolation to fill in a few broken data points in the collected data. The linear interpolation method is as follows:

[0074]

[0075] Where N is the number of missing points, Y n For the data value of the nth missing point, Y0 and Y... N+1 These are the data values ​​immediately before and after the missing sequence.

[0076] S103: Establish a first-order thermal dynamic gray box model of the building.

[0077] The above-mentioned first-order thermal dynamic gray box model of the building is a 2R1C model, which includes two R-term parameters and one C-term parameter. The model is as follows:

[0078]

[0079] Where C is the building's total lumped heat capacity, is the C-parameter of the building's first-order thermal dynamic gray box model, and R... opaque and R window Let T be the thermal resistance of the building envelope, R be the parameter of the first-order thermal dynamic gray box model of the building, and T be the thermal resistance of the building envelope. in Indoor temperature, T out For the outdoor temperature, the optimization range for parameters C and R is set to positive real numbers only, without setting an upper limit.

[0080] S105: Construct the fitness function of the particle swarm algorithm based on the accuracy evaluation index of the first-order thermal dynamic gray box model of the building.

[0081] The accuracy evaluation index of the above-mentioned first-order thermal dynamic gray box model of the building takes the minimum absolute value of the difference between the predicted indoor temperature and the actual indoor temperature as the objective function. The objective function is:

[0082]

[0083] Where n is the sample number and N is the sample size. This represents the actual indoor temperature. This represents the predicted indoor temperature.

[0084] S107: Solve the parameters of the first-order thermal dynamic gray box model of a building using the particle swarm optimization algorithm.

[0085] The particle swarm optimization algorithm described above includes the following steps:

[0086] S1071: Set the built-in parameters of the particle swarm optimization algorithm, including particle population size, maximum number of iterations, learning factor and external solution set size;

[0087] S1072: Randomly initialize the velocity and position of the particle swarm;

[0088] S1073: Substitute the particles into the first-order thermal dynamic gray box model of the building, calculate the fitness function of each particle based on the fitness function of the particle swarm algorithm, and obtain the fitness function value of the particle.

[0089] S1074: Determine the relationship between the current fitness value of each particle and its individual best value pbest. If the fitness value of this generation of particles is superior, do not update it; otherwise, update the individual best value pbest. Compare pbest with the global best value gbest. If pbest is superior, update the gbest value.

[0090] S1075: Update the next generation of particles according to the particle renewal rate formula;

[0091] S1076: Repeat steps S1073 to S1075 until the set minimum error is met or the maximum number of iterations is reached, to obtain the parameter values ​​of the first-order thermal dynamic gray box model of the building.

[0092] In step S1075 above, the formula for the particle renewal rate is:

[0093]

[0094]

[0095] in, It is the d-th dimension component of the velocity vector of particle i in the k-th iteration. pbest is the d-th dimension component of the position of particle i in the k-th iteration.id It is the d-th dimension component of the historical individual optimal value position of particle i, gbest id ω is the d-th dimension component of the historical global optimal position of all particles, c1 and c2 are acceleration constants, and r1 and r2 are random functions.

[0096] S109: Establish a second-order thermal dynamic gray box model of the building and set the parameter range of the second-order thermal dynamic gray box model of the building.

[0097] The above-mentioned second-order thermal dynamic gray box model of the building is a 3R2C model, which includes three R parameters and two C parameters. The specific model is as follows:

[0098]

[0099]

[0100] When setting the optimization range of the R and C parameters in the second-order thermal dynamic gray box model of the building, the R and C parameters in the first-order thermal dynamic gray box model of the building are used. The specific optimization range of the R and C parameters in the above-mentioned second-order thermal dynamic gray box model of the building is as follows:

[0101]

[0102] C air <10 log(C)+1 C wall >10 log(C)+1

[0103] Wherein, parameter C air and C wall For parameter C in the second-order thermal dynamic gray box model of the building, C air For the building air heat capacity, C wall For the total heat capacity of the building envelope, parameters R1 and R2 are given. wall,outer and R wall,inner R is the parameter in the second-order thermal dynamic gray box model of the building, where R1 is the thermal resistance term directly related to the indoor and outdoor dry-bulb temperatures. wall,inner R represents the overall thermal resistance of the building's integrated building envelope. wall,outer For the overall thermal resistance of the building envelope on the outside, T wall Q represents the temperature of the building's lumped envelope. other Other loads, such as those caused by heat generated by electrical appliances.

[0104] S111: Solve for the parameters of the second-order thermal dynamic gray box model of a building using the particle swarm optimization algorithm.

[0105] S113: Calculate the building load based on the solution results of the second-order thermal dynamic gray box model parameters of the building.

[0106] Building load is calculated in the following way:

[0107]

[0108] Among them, Q building For building load.

[0109] This invention provides a building load prediction method based on the operating data of a multi-split air conditioning system. Addressing the problem that existing techniques cannot stably identify building thermal dynamic parameters using the high-frequency, wide-range oscillation output values ​​of direct expansion air conditioning systems, leading to local optima and rendering the identification model unusable during normal operations or causing it to fail during online model updates, this invention proposes a two-stage gray-box model identification method for building thermal dynamics. This method ensures that building thermal dynamic models built based on the capabilities of direct expansion air conditioning systems can be well identified and maintain good identification stability, providing necessary support for the online model update process. Furthermore, this invention, based on the identification domain in the parameter identification optimization process, applies more reasonable feasible region constraints during the optimization problem establishment process and before the optimization begins. These constraints are based on a simpler, first-order model, thus exhibiting high robustness. This prevents variables in the parameter optimization process from falling into local optima outside the feasible region during random processes, ensuring good identification stability in the building thermal dynamics model identification problem.

[0110] Compared with the prior art, the embodiments of the present invention can bring the following technical effects:

[0111] 1) Stable building thermal dynamic model parameter identification results can be obtained based on direct expansion air conditioning capacity data.

[0112] 2) This invention can be universally applied to the establishment of thermal dynamic models of any building equipped with a direct expansion air conditioning system, and solves the problem that the fluctuation capability of the direct expansion air conditioning system cannot be accurately identified by automated parameters.

[0113] The present invention will now be described in detail with reference to preferred embodiments.

[0114] This invention provides a building load prediction method based on VRF (Variable Refrigerant Flow) air conditioning system operation data. By establishing a second-order gray box mathematical model of the building and constraining the optimization range of the second-order model parameters through the parameter identification results of the first-order gray box model, the method uses particle swarm optimization to perform optimization. This method can provide a way to construct a thermal dynamic model of a building under the operation data of a multi-split air conditioning system, thereby improving the feasibility and accuracy of building load prediction.

[0115] like Figure 2As shown, a building load forecasting method based on VRF air conditioning system operation data provided by the present invention includes:

[0116] Step 1: Establish a multi-unit energy efficiency data calculation model.

[0117] The multi-split air conditioner energy efficiency calculation model obtains the total flow rate through compressor frequency calculation, and then calculates the final flow rate to the indoor unit through each bypass valve. Based on the refrigerant temperature at the outlet of the main pipe of the subcooling heat exchanger and the average temperature and pressure at the outlet of the evaporator, the refrigerant enthalpy difference is calculated to complete the multi-split air conditioner energy efficiency data calculation.

[0118] The calculation model for the energy efficiency data of multi-split air conditioners is as follows:

[0119] Q VRF =F idu (h in -h out ) / 1000

[0120] Among them, Q VRF The cooling capacity of a multi-split indoor unit is measured in watts (W / h). in h out These are the enthalpy values ​​of the refrigerant entering and exiting the indoor unit, respectively, in kJ / kg; F idu The refrigerant flow rate entering the indoor unit of a multi-split air unit, in kg / s, can be expressed by the following formula:

[0121]

[0122] Among them, F comp F represents the refrigerant mass flow rate at the compressor outlet, expressed in kg / s. bypass,i The refrigerant mass flow rate of bypass loop i is expressed in kg / s, and n is the number of bypass loops. The flow rate calculation of the bypass loops is related to the corresponding valve opening status and the opening degree of the electronic expansion valve. The calculated refrigeration capacity value of the multi-unit system is as follows: Figure 4 As shown.

[0123] When collecting data, the data collection interval for the cooling capacity of the multi-unit system is 30 seconds. Other collected data include the average return air temperature of the indoor unit and the dry bulb temperature of the outdoor unit fan inlet.

[0124] The collected data needs to be preprocessed, including data cleaning and data completion. The main purpose of data cleaning is to remove outlier data, including meaningless negative sensor values ​​and transmission anomalies. Transmission anomaly data is filtered using a threshold set based on the rated capacity of the multi-split air conditioner; in this example, the cooling capacity of the multi-split air conditioner is on the order of 10. 5 Set the upper limit of the threshold to 10. 6 The temperature value is on the order of 10, and the upper limit of the threshold is set to 50℃.

[0125] Data completion uses linear interpolation. For a small number of breakpoints in the data, the calculation formula is as follows:

[0126]

[0127] N (N≤5) is the number of missing points, Y n (n = 1, 2, ..., N) represents the data value of the nth missing point, Y0 and Y... N+1 These are the data immediately before and after the missing sequence. In cases of long-term missing data, the labeled data is abnormal and will not be used in subsequent training.

[0128] Step 2: Establish the mathematical expression for the 2R1C model of the building.

[0129] The mathematical description of the first-order thermal dynamic model of the building in this preferred embodiment is as follows:

[0130]

[0131] Where C represents the building's total lumped heat capacity in J / K, is the C-parameter of the building's first-order thermal dynamic gray box model, and R... opaque and R window The thermal resistance of the building envelope is expressed in K / W, and T is the R-parameter of the first-order thermal dynamic gray box model of the building. in Indoor temperature, in °C (°C), T out Outdoor temperature, in °C, Q VRF The real-time cooling capacity of the VRF system is expressed in W. For the parameters C and R, only the optimization range is set to positive real numbers, without setting an upper limit.

[0132] For a first-order building model, there is no need to impose upper limit constraints on C and R; only the optimization range is set to positive real numbers.

[0133] Step 3: Construct the particle swarm optimization fitness function with the objective function of minimizing the absolute value of the difference between the model-calculated indoor temperature and the actual indoor temperature.

[0134] The objective function is set as follows:

[0135]

[0136] Where n is the sample number and N is the sample size. This refers to the actual indoor temperature, in °C. Indoor temperature is predicted in °C.

[0137] Step 4: Use the particle swarm optimization algorithm to solve for the optimal model parameters of the cold building 2R1C model.

[0138] The specific implementation steps based on the particle swarm optimization algorithm are as follows:

[0139] (1) Set the built-in parameters of the algorithm, including particle population size, maximum number of iterations, learning factor, and external solution set size;

[0140] (2) Randomly initialize the velocity and position of the particle swarm;

[0141] (3) Substitute the particles into the above-mentioned first-order thermal dynamic gray box model of the building, and calculate the fitness function of each particle based on the fitness function of the particle swarm algorithm to obtain the fitness function value of the particle.

[0142] (4) Determine the relationship between the current fitness value of each particle and the individual optimal value pbest. If the fitness value of this generation of particles is superior, do not update; otherwise, update the individual optimal value pbest. Compare pbest with the global optimal value gbest. If pbest is superior, update the gbest value.

[0143] The next generation of particles is updated according to the particle renewal rate formula:

[0144]

[0145]

[0146] in, It is the d-th component of the velocity vector of particle i in the k-th iteration; It is the d-th dimension component of the position of particle i in the k-th iteration; pbest id It is the d-th dimension component of the historical individual optimal value position of particle i; gbest id ω is the d-th dimension component of the historical global optimal position of all particles; ω is the inertial weight, which is non-negative; c1 and c2 are acceleration constants; r1 and r2 are random functions with values ​​in the range [0, 1].

[0147] Repeat steps (3) and (4) until the set minimum error is met or the maximum number of iterations is reached. Stop the search and output the external archive, which is the parameter value of the building 2R1C gray box model: C, R opaque R window .

[0148] Step 5: Establish the mathematical expression for the 3R2C model of the building.

[0149] In this preferred embodiment, the mathematical description of the second-order thermal dynamics model of the building is as follows:

[0150]

[0151]

[0152] Among them, C air C wall These are the building air heat capacity and the building envelope heat capacity, respectively, in J / K and T. out T in T wall These represent outdoor temperature, indoor temperature, and building lumped wall temperature, respectively, in °C; R1 is the thermal resistance term directly connected to the indoor and outdoor dry-bulb temperatures, in K / W; R wall,inner R wall,outer Q represents the combined thermal resistance of the inner and outer sides of the building envelope, respectively, in K / W; VRF Real-time cooling capacity of the VRF system, in W and Q. other Other loads, such as those caused by heat generated by electrical appliances.

[0153] Step Six: Based on the obtained C and R... opaque and R window Set the parameter optimization range for the building 3R2C model.

[0154] The parameter optimization range of the above-mentioned 3R2C building model includes the R-term parameters and the C-term parameters. The R-term optimization range in the 3R2C building model is as follows:

[0155]

[0156] The optimization range for term C in the 3R2C building model is as follows:

[0157] C air <10 log(C)+1

[0158] C wall >10 log(C)+1

[0159] Step 7: Combining steps 3 and 4, calculate the building model parameter C under the 3R2C model. air C wall R1, R2 and R3.

[0160] In this preferred embodiment, the identification results of the parameter identification method without the additional constraints in step six on the same dataset were compared, as shown in Table 1.

[0161] Table 1. Parameter identification results in the 3R2C building model.

[0162]

[0163] In this preferred embodiment, the temperature identification results of the model without the additional constraints in step S6 were compared on the same dataset. Figure 5 As shown.

[0164] Step 8: Based on the parameter identification results of the desired building 3R2C model, calculate the building load.

[0165] Specifically, the formula for calculating building load is as follows:

[0166]

[0167] Among them, C air Building air heat capacity, unit: J / K; T out T in T wall These represent outdoor temperature, indoor temperature, and building lumped wall temperature, respectively, in °C; R1 is the thermal resistance term directly connected to the indoor and outdoor dry-bulb temperatures, in K / W; R wall,inner Q represents the overall thermal resistance of the building's integrated building envelope, expressed in K / W. building This represents the building's real-time heat load, measured in W.

[0168] In this specific embodiment, the invention is implemented based on the technical solution of the present invention. Utilizing historical data from a multi-split air conditioning system in a building in Foshan City, Guangdong Province, and an open-source Python library, a detailed implementation method and specific operation process are provided, along with applicable parameters. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the scope of protection of the present invention. Those skilled in the art understand that, in addition to implementing the system, device, and its modules provided by the present invention in purely computer-readable program code, the same program can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps.

[0169] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A building load forecasting method based on multi-split air conditioning system operation data, characterized in that, The method includes the following steps: S101: Establish a calculation model for the energy efficiency data of multi-split air conditioners and obtain energy efficiency data of multi-split air conditioners; S103: Establish a first-order thermal dynamic gray box model of the building; S105: Construct the particle swarm optimization algorithm fitness function based on the accuracy evaluation index of the first-order thermal dynamic gray box model of the building. S107: Using the particle swarm optimization algorithm, solve for the parameters of the first-order thermal dynamic gray box model of the building; S109: Establish a second-order thermal dynamic gray box model of the building and set the parameter range of the second-order thermal dynamic gray box model of the building; S111: Using the particle swarm optimization algorithm, solve for the parameters of the second-order thermal dynamic gray box model of the building; S113: Calculate the building load based on the solution results of the second-order thermal dynamic gray box model parameters of the building; in, In step S101, the energy efficiency data calculation model for the multi-split air conditioner is as follows: It is calculated using the following formula: in, To enhance the cooling capacity of the multi-split system in the indoor unit, For the refrigerant flow rate entering the multi-split indoor unit, The enthalpy of the refrigerant entering the indoor unit. The enthalpy of the refrigerant flowing out of the indoor unit. This refers to the refrigerant mass flow rate at the compressor outlet. For bypass circuit i Refrigerant mass flow rate n This represents the number of bypass loops. The first-order thermal dynamic gray box model of the building is as follows: Where C represents the building's total lumped heat capacity, and represents the C-parameter of the building's first-order thermal dynamic gray box model. and Let be the thermal resistance of the building envelope, and be the R-parameter of the first-order thermal dynamic gray box model of the building. Indoor temperature, For the outdoor temperature, the optimization range for parameters C and R is set to positive real numbers only, without setting an upper limit; In step S109, the second-order thermal dynamic gray box model of the building is: Based on the R and C parameters in the first-order thermal dynamic gray box model of the building, the optimization range of the R and C parameters in the second-order thermal dynamic gray box model of the building is set as follows: Among them, parameters and For parameter C in the second-order thermal dynamic gray box model of the building, For building air heat capacity, For the heat capacity of the building's lumped envelope, parameters , and The R-term parameter in the second-order thermal dynamic gray box model of the building is... This is the thermal resistance term, which is directly connected to the indoor and outdoor dry-bulb temperatures. The overall thermal resistance of the building's integrated building envelope. The overall thermal resistance of the building envelope on the outside is considered. For the temperature of the building's overall building envelope, For other loads; In step S113, the building load is calculated as follows: in, For building load.

2. The method as described in claim 1, characterized in that, Step S101 further includes the collection and preprocessing of operating data from the multi-split air conditioning system. The preprocessing includes data cleaning and data completion. The data cleaning is set to remove abnormal data, which includes negative sensor data and abnormal transmission data. The abnormal transmission data is filtered based on the rated capacity of the multi-split air conditioner. The data completion uses a linear interpolation method to complete a small number of broken data points in the collected data. The linear interpolation method is as follows: in, This represents the number of missing points. For the first n The data values ​​of the missing points and These are the data values ​​immediately before and after the missing sequence.

3. The method as described in claim 2, characterized in that, The data acquisition period for the multi-unit refrigeration capacity is 30 seconds, and the upper limit of the filtering threshold for the multi-unit refrigeration capacity value is 10. 6 The upper limit of the screening threshold for the temperature value is 50℃.

4. The method as described in claim 1, characterized in that, In step S105, the accuracy evaluation index of the first-order thermal dynamic gray box model of the building takes the minimum absolute value of the difference between the predicted indoor temperature and the actual indoor temperature as the objective function. The objective function is: in, n For sample number, N For the sample size, This represents the actual indoor temperature. This represents the predicted indoor temperature.

5. The method as described in claim 4, characterized in that, In step S107, the particle swarm optimization algorithm includes the following steps: S1071: Set the built-in parameters of the particle swarm optimization algorithm, including particle population size, maximum number of iterations, learning factor and external solution set size; S1072: Randomly initialize the velocity and position of the particle swarm; S1073: Substitute the particles into the first-order thermal dynamic gray box model of the building, calculate the fitness function of each particle based on the fitness function of the particle swarm algorithm, and obtain the fitness function value of the particle. S1074: Determine the relationship between the current fitness value of each particle and its individual best value pbest. If the fitness value of this generation of particles is superior, do not update it; otherwise, update the individual best value pbest. Compare pbest with the global best value gbest. If pbest is superior, update the gbest value. S1075: Update the next generation of particles according to the particle renewal rate formula; S1076: Repeat steps S1073 to S1075 until the set minimum error is met or the maximum number of iterations is reached, to obtain the parameter values ​​of the first-order thermal dynamic gray box model of the building.

6. The method as described in claim 5, characterized in that, The particle update rate formula in the particle swarm optimization algorithm is: in, It is the d-th dimension component of the velocity vector of particle i in the k-th iteration. It is the d-th dimension component of the position of particle i in the k-th iteration. It is the d-th dimension component of the historical individual optimal value position of particle i. It is the d-th dimension component of the historical global optimal position of all particles. For inertial weights, and Let be the acceleration constant. and It is a random function.

Images