Air conditioner control method and system

Abstract

The invention relates to the technical field of refrigeration equipment control, in particular to an air conditioner control method and system. Firstly, a refrigerant flow prediction model is established by combining the advantages of LSTM and XGBoost, a time sequence and static characteristics are accurately captured, prediction precision and system adaptability are improved, reliable data support is provided for intelligent regulation and control, refrigerant distribution is optimized, energy efficiency is improved, and energy consumption is reduced; secondly, the opening degree of an electronic expansion valve and the frequency of a compressor are controlled in a feedforward mode through a multivariable cooperative control algorithm, the load and the flow are accurately matched, the system efficiency is improved, it is ensured that all assemblies operate in a coordinated mode, energy waste is reduced, and stability and comfort are optimized; finally, PID feedback control is combined, the frequencies of the electronic expansion valve and the compressor are finely adjusted in real time, the optimal energy efficiency and stability of the system are ensured, closed-loop control is achieved, the temperature control precision is further improved, fluctuation is reduced, and the energy efficiency is optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of refrigeration equipment control, and in particular to an air conditioning control method and system. Background Art

[0002] Multi-split air conditioners are widely used in large commercial buildings, hotels, office buildings and other places. Due to their efficient cooling and heating capabilities and flexible installation methods, they have gradually become the mainstream choice of modern air conditioning technology. This type of system can independently adjust the working state of each indoor unit according to the temperature requirements of different areas, thereby achieving a high degree of energy efficiency optimization and comfort control. Efficient multi-split air conditioning control systems not only help save energy costs, but also reduce environmental pollution and improve the overall comfort of buildings, becoming a core component of smart buildings and green environmental protection concepts.

[0003] However, the existing multi-split air-conditioning control system still has certain technical difficulties in practical applications. Traditional control methods rely on simple PID control or fixed preset parameters, which are difficult to cope with dynamic changes in system load and complex environmental changes. Due to the inability to accurately predict and adjust the required flow of each indoor unit, the refrigerant is unevenly distributed, resulting in low energy efficiency, poor temperature control accuracy, frequent system start and stop, and even overheating or overcooling, which affects the stability and comfort of the system. In addition, the traditional control method lacks intelligent adaptive capabilities, making it difficult to achieve refined control and effectively reduce energy consumption. Therefore, it is necessary to develop an air-conditioning control system that is more intelligent, accurate, and can adapt to complex working conditions.

[0004] To this end, an air conditioning control method and system are proposed. Summary of the invention

[0005] The present invention provides an air conditioning control method and system. By combining the LSTM and XGBoost refrigerant flow prediction model, the prediction accuracy and system adaptability are improved, the refrigerant distribution is optimized, the energy efficiency is improved and the energy consumption is reduced; a multivariable collaborative control algorithm is used to accurately adjust the electronic expansion valve opening and the compressor frequency to ensure that the load matches the flow, improve the system efficiency and reduce energy waste; at the same time, PID feedback control adjusts the system parameters in real time to achieve closed-loop control, improve temperature control accuracy and overall energy efficiency.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An air conditioning control method, comprising:

[0008] Real-time collection of status parameters of each indoor unit in the multi-split system; the status parameters include evaporator inlet and outlet temperature and pressure, condenser inlet and outlet temperature and pressure, ambient temperature and humidity, and current opening of the electronic expansion valve;

[0009] Based on the state parameters, the required flow of each indoor unit is calculated by a pre-trained refrigerant flow prediction model, and a global refrigerant allocation strategy is generated in combination with the total system load;

[0010] According to the global refrigerant distribution strategy, a multivariable collaborative control algorithm is used to feedforward control the opening of the electronic expansion valve and synchronously adjust the compressor frequency to match the system load with the flow rate;

[0011] Monitor the evaporator outlet superheat deviation and compressor frequency deviation, and control the electronic expansion valve opening and compressor frequency through PID controller feedback.

[0012] Furthermore, the establishment of the refrigerant flow prediction model includes:

[0013] Acquire historical data records of the indoor unit, and preprocess the historical data records to obtain preprocessed data records; the historical data records include historical state parameters and historical demand flow;

[0014] The preprocessed data records are divided into a training set and a validation set in a ratio of 7:3, the training set is used to train the fusion model of LSTM combined with XGBoost, and the fusion model is optimized through a loss function;

[0015] When the loss function converges, the fusion model is verified using a verification set; if the verification error is greater than an error threshold, the fusion model is retrained; otherwise, a refrigerant flow prediction model is obtained.

[0016] Furthermore, the refrigerant flow prediction model includes:

[0017] Input layer, used to receive the status parameters of the indoor unit;

[0018] A fusion layer, including LSTM units and XGBoost units, is used to extract and integrate the time series features and static features of the state parameters;

[0019] Prediction layer, used to output prediction results;

[0020] The error feedback layer is used to execute the loss function and reduce the training error during model training;

[0021] The output layer is used to output the required flow.

[0022] Furthermore, the global refrigerant allocation strategy generated in combination with the total system load includes:

[0023] Calculate total system load and compressor target frequency;

[0024] Establishing a global refrigerant allocation constraint function according to the real-time monitoring data and the compressor target frequency; the real-time monitoring data includes the set temperature, real-time temperature, real-time flow rate and required flow rate of the indoor unit;

[0025] Solve the global refrigerant allocation constraint function to obtain a global refrigerant allocation strategy; the calculation formula of the global refrigerant allocation constraint function is:

[0026]

[0027] Among them, min represents the minimization function, i represents the index number of the indoor unit, T set,i represents the set temperature of the i-th indoor unit, T real,i represents the real-time temperature of the i-th indoor unit, m i represents the real-time flow of the i-th indoor unit, m pre,i represents the demand flow of the i-th indoor unit, f comp Indicates the compressor target frequency; ω 1 ,ω 2 and ω 3 They represent temperature weight, flow weight and load weight respectively.

[0028] Furthermore, the feedforward control includes:

[0029] Defining system state variables and system control variables, and establishing state equations according to the system state variables and the system control variables;

[0030] Solving the state equation to obtain the electronic expansion valve opening adjustment amount and the compressor frequency adjustment amount;

[0031] The electronic expansion valve opening and the compressor frequency are adjusted according to the electronic expansion valve opening adjustment amount and the compressor frequency adjustment amount.

[0032] Furthermore, the feedback control includes:

[0033] Calculate the superheat and superheat deviation at the evaporator outlet;

[0034] Adjust the opening of the electronic expansion valve based on PID control to match the refrigerant flow to the load demand and ensure the stability of superheat;

[0035] Adjust the compressor frequency based on the electronic expansion valve opening feedback to ensure that the refrigerant supply matches the system load demand and avoid frequent start and stop of the compressor;

[0036] Continuously monitor system status and optimize gain coefficients.

[0037] The present invention also provides an air conditioning control system, comprising:

[0038] A data acquisition module is used to collect the status parameters of each indoor unit in the multi-split system in real time; the status parameters include the inlet and outlet temperature and pressure of the evaporator, the inlet and outlet temperature and pressure of the condenser, the ambient temperature and humidity, and the current opening of the electronic expansion valve;

[0039] A refrigerant flow prediction module, used to calculate the required flow of each indoor unit based on the state parameters through a pre-trained refrigerant flow prediction model, and generate a global refrigerant allocation strategy in combination with the total system load;

[0040] A feedforward control module, configured to feedforward control the opening of the electronic expansion valve using a multivariable collaborative control algorithm according to the global refrigerant distribution strategy, and synchronously adjust the compressor frequency to match the system load with the flow rate;

[0041] The feedback control module is used to monitor the evaporator outlet superheat deviation and compressor frequency deviation, and control the electronic expansion valve opening and compressor frequency through PID controller feedback.

[0042] The beneficial effects of the present invention are:

[0043] 1. The present invention proposes a refrigerant flow prediction model, which, by combining the advantages of LSTM and XGBoost, can simultaneously capture the time series characteristics and static characteristics of the refrigerant flow, thereby improving the prediction accuracy and system adaptability, and can more accurately predict the demand flow of each indoor unit, providing reliable data support for the intelligent control of the air-conditioning system, optimizing refrigerant distribution, improving overall energy efficiency and reducing energy consumption.

[0044] 2. The present invention uses a multivariable collaborative control algorithm to feedforward control the opening of the electronic expansion valve and synchronously adjust the compressor frequency. This method can accurately match the system load with the refrigerant flow demand, thereby achieving efficient operation of the system. This strategy not only improves refrigeration efficiency and reduces energy consumption, but also ensures the coordinated work of various components, reduces unnecessary energy waste, and optimizes the stability and comfort of the air-conditioning system, improving the overall system performance.

[0045] 3. The present invention monitors the deviation of superheat at the evaporator outlet and the deviation of the compressor frequency, and combines the feedback adjustment of the PID controller to fine-tune the opening of the electronic expansion valve and the compressor frequency in real time, thereby ensuring that the system maintains the best energy efficiency and stability during operation. Combined with feedforward control, the PID feedback mechanism provides more sophisticated dynamic adjustment, enabling the multi-split air conditioning system to achieve closed-loop control, further improving temperature control accuracy, reducing fluctuations, and optimizing refrigerant distribution, ensuring accurate matching of load and flow, and enhancing the system's responsiveness and overall energy efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0047] Figure 1 This is a flow chart of an air conditioning control method provided by the present invention;

[0048] Figure 2 It is a structural schematic diagram of a refrigerant flow prediction model provided by the present invention;

[0049] Figure 3 This is a structural diagram of an air conditioning control system provided by the present invention. DETAILED DESCRIPTION

[0050] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0051] Embodiment 1

[0052] An air conditioning control method, such as Figure 1 As shown, including:

[0053] S100: Real-time collection of status parameters of each indoor unit in the multi-split system; the status parameters include evaporator inlet and outlet temperature and pressure, condenser inlet and outlet temperature and pressure, ambient temperature and humidity, and current opening of the electronic expansion valve;

[0054] S200: Based on the state parameters, the required flow of each indoor unit is calculated by a pre-trained refrigerant flow prediction model, and a global refrigerant allocation strategy is generated in combination with the total system load;

[0055] Furthermore, the establishment of the refrigerant flow prediction model includes:

[0056] Acquire historical data records of the indoor unit, and preprocess the historical data records to obtain preprocessed data records; the historical data records include historical state parameters and historical demand flow;

[0057] The preprocessed data records are divided into a training set and a validation set in a ratio of 7:3, the training set is used to train the fusion model of LSTM combined with XGBoost, and the fusion model is optimized through a loss function;

[0058] When the loss function converges, the fusion model is verified using a verification set; if the verification error is greater than an error threshold, the fusion model is retrained; otherwise, a refrigerant flow prediction model is obtained.

[0059] Specifically, the preprocessing process includes outlier detection, missing value filling and data normalization, among which outlier detection is used to remove sensor error data (such as negative temperature, unreasonable EEV opening), missing value filling is used to fill missing data using interpolation or KNN, and data normalization is used to normalize data of different dimensions to improve the model convergence speed.

[0060] Furthermore, the refrigerant flow prediction model is as follows: Figure 2 As shown, including:

[0061] Input layer, used to receive the status parameters of the indoor unit;

[0062] A fusion layer, including LSTM units and XGBoost units, is used to extract and integrate the time series features and static features of the state parameters;

[0063] Prediction layer, used to output prediction results;

[0064] The error feedback layer is used to execute the loss function and reduce the training error during model training;

[0065] The output layer is used to output the required flow.

[0066] Specifically, the input layer is used to receive the state parameters of the indoor unit; the fusion layer includes an LSTM unit and an XGBoost unit, wherein the LSTM unit includes three LSTM sub-units connected in series, which are used to extract the time series features of the state parameters, and the XGBoost unit includes an XGBoost sub-unit (excluding the prediction part), which is used to extract the static features of the state parameters. The time series features and the static features are fused in the fusion layer through the SUM function; the prediction layer includes a RELU activation function and a fully connected neural network, the RELU activation function is used to improve the nonlinear fitting ability of the fusion features, and the fully connected neural network receives the activated fusion features and outputs the prediction results; the error feedback layer is used to execute the loss function. In this embodiment, the loss function is not limited, and is preferably the mean square error; the output layer is used to output the demand flow. When the loss value of the loss function in 10 epochs is less than the loss threshold, the demand flow is output. In this embodiment, the loss threshold is preferably 0.17.

[0067] The model combines the advantages of LSTM and XGBoost, and can capture the time series characteristics and static characteristics of refrigerant flow at the same time, improving the prediction accuracy. Through the optimization of the loss function of the error feedback layer, the model can adaptively adjust parameters, reduce training errors, and enhance generalization capabilities. The use of the refrigerant flow prediction model ensures that after deep feature extraction and error optimization of the input data, high-precision demand flow prediction results are finally output, providing accurate data support for the intelligent control of the air-conditioning system, improving energy efficiency and operational stability.

[0068] Furthermore, the global refrigerant allocation strategy generated in combination with the total system load includes:

[0069] Calculate total system load and compressor target frequency;

[0070] Establishing a global refrigerant allocation constraint function according to the real-time monitoring data and the compressor target frequency; the real-time monitoring data includes the set temperature, real-time temperature, real-time flow rate and required flow rate of the indoor unit;

[0071] Solve the global refrigerant allocation constraint function to obtain a global refrigerant allocation strategy; the calculation formula of the global refrigerant allocation constraint function is:

[0072]

[0073] Among them, min represents the minimization function, i represents the index number of the indoor unit, T set,i represents the set temperature of the i-th indoor unit, T real,i represents the real-time temperature of the i-th indoor unit, m i represents the real-time flow of the i-th indoor unit, m pre,i represents the demand flow of the i-th indoor unit, f comp Indicates the compressor target frequency; ω 1 ,ω 2 and ω 3 They represent temperature weight, flow weight and frequency weight respectively.

[0074] Specifically, the total system load is calculated as:

[0075]

[0076] Among them, Q total Indicates the total system load, i indicates the index number of the indoor unit, T set,i represents the set temperature of the i-th indoor unit, T real,i represents the real-time temperature of the i-th indoor unit, C p is the specific heat capacity of air, V air Indicates the air mass flow rate passing through the indoor unit heat exchanger per unit time; the calculation formula for the compressor target frequency is:

[0077]

[0078] Among them, f comp represents the compressor target frequency, f min Indicates the minimum frequency of the compressor, f max Indicates the maximum frequency of the compressor, Q total Indicates the total system load, Q raterepresents the rated load of the system; a global refrigerant allocation constraint function is established according to the real-time monitoring data and the compressor target frequency, wherein the real-time monitoring data includes the set temperature, real-time temperature, real-time flow rate and demand flow rate of the indoor unit; the global refrigerant allocation constraint function is solved to obtain the global refrigerant allocation strategy.

[0079] By calculating the total system load and the target frequency of the compressor, combined with real-time monitoring data, a global refrigerant allocation constraint function is constructed, and the refrigerant allocation strategy is optimized by solving the function. This strategy can achieve coordinated adjustment of the electronic expansion valve and the compressor while ensuring the cooling demand of each indoor unit, so that the system load and the refrigerant flow are dynamically matched, thereby improving the energy efficiency ratio, reducing energy consumption, and improving the overall stability and comfort of the air-conditioning system.

[0080] S300: According to the global refrigerant distribution strategy, a multivariable collaborative control algorithm is used to feedforward control the opening of the electronic expansion valve, and the compressor frequency is synchronously adjusted to match the system load with the flow rate;

[0081] Furthermore, the feedforward control includes:

[0082] Defining system state variables and system control variables, and establishing state equations according to the system state variables and the system control variables;

[0083] Solving the state equation to obtain the electronic expansion valve opening adjustment amount and the compressor frequency adjustment amount;

[0084] The electronic expansion valve opening and the compressor frequency are adjusted according to the electronic expansion valve opening adjustment amount and the compressor frequency adjustment amount.

[0085] Specifically, the system state variables and system control variables are defined, where the system state variable X = [e T,1 ,e T,2 ,...,e T,n ,Δm 1 ,Δm 2 ,...,Δm n ,f comp ] T , where e T Indicates room temperature deviation, calculated by e T =T set,i -T real,i , Δm represents the change in refrigerant flow rate, and the calculation formula is Δm=m i -m pre,i ; System control variable U = [O EEV,1 ,O EEV,2 ,...,O EEV,n ,f comp ] T , where OEEV Indicates the opening degree of the electronic expansion valve; the calculation formula of the state equation is:

[0086] X(k+1)=AX(k)+BU(k);

[0087] Wherein, k represents the current state, k+1 represents the next state, A represents the system state matrix, which represents the dynamic relationship between temperature, flow rate and compressor frequency, and B represents the control input matrix, which describes the influence of the opening of the electronic expansion valve and the compressor frequency on the system state. A and B can be obtained based on the physical modeling method and the data-driven method. The physical modeling method establishes the state equation based on thermodynamics and fluid mechanics, and the data-driven method combines the LSTM training data-driven model and extracts the approximate state matrix therefrom. In this embodiment, the data-driven method is used to obtain, and the process can be expressed as follows:

[0088] X(k+1)=f LSTM (X(k),U(k));

[0089] Among them, f LSTM The LSTM model is represented. The LSTM is used to train the state transfer relationship. After the training is completed, the partial derivative of each variable is calculated from the LSTM to extract the approximate values ​​of A and B. The state equation is solved to obtain the electronic expansion valve opening adjustment amount and the compressor frequency adjustment amount. The electronic expansion valve opening and the compressor frequency are adjusted according to the electronic expansion valve opening adjustment amount and the compressor frequency adjustment amount.

[0090] By defining system state variables and control variables, establishing and solving state equations, the adjustment of the electronic expansion valve opening and compressor frequency can be made more accurate and efficient. This optimization strategy can ensure that the refrigerant flow rate is dynamically matched with the system load, improve the control response speed, reduce control deviation, and effectively avoid problems such as frequent compressor start and stop and insufficient refrigerant supply, thereby improving the operating stability, energy efficiency and overall comfort of the air-conditioning system.

[0091] S400: Monitor the evaporator outlet superheat deviation and compressor frequency deviation, and control the electronic expansion valve opening and compressor frequency through PID controller feedback.

[0092] Furthermore, the feedback control includes:

[0093] Calculate the superheat and superheat deviation at the evaporator outlet;

[0094] Adjust the opening of the electronic expansion valve based on PID control to match the refrigerant flow to the load demand and ensure the stability of superheat;

[0095] Adjust the compressor frequency based on the electronic expansion valve opening feedback to ensure that the refrigerant supply matches the system load demand and avoid frequent start and stop of the compressor;

[0096] Continuously monitor system status and optimize gain coefficients.

[0097] Specifically, the superheat SH i =T evap,i -T sat (P evap,i ), where T evap,i represents the outlet temperature of the evaporator of the i-th indoor unit, P evap,i represents the outlet pressure of the evaporator of the i-th indoor unit, T sat (P evap,i ) indicates that at pressure P evap,i Saturation temperature under the state, superheat deviation ΔSH i =SH i -SH set , where SH set Indicates setting the superheat; based on PID control, the electronic expansion valve opening is adjusted to match the refrigerant flow with the load demand to ensure a stable superheat. The PID calculation formula is:

[0098]

[0099] Among them, K p , K i and K d Represents the gain coefficient of the PID controller; the compressor frequency is adjusted according to the feedback of the electronic expansion valve opening, and the calculation formula for adjusting the compressor frequency can be:

[0100] f comp (k+1)=f comp (k)+K f ·∑(O EEV,i (k+1)-O EEV,i (k));

[0101] Among them, K f Express the compressor frequency adjustment gain coefficient; continuously monitor the system status and optimize the gain coefficient, wherein the gain coefficient includes the gain coefficient of the PID controller and the compressor frequency adjustment gain coefficient.

[0102] By calculating the superheat and its deviation at the evaporator outlet in real time, the electronic expansion valve opening is accurately adjusted using PID control to dynamically match the refrigerant flow to the load demand and ensure the stability of the superheat. At the same time, the compressor frequency is adjusted based on the feedback of the electronic expansion valve opening to ensure that the refrigerant supply matches the system load and avoid increased energy consumption and equipment loss caused by frequent start and stop of the compressor. By continuously monitoring the system status and optimizing the gain coefficient, the system has stronger adaptive capabilities, further improving the operating stability, energy efficiency ratio and overall control accuracy of the air-conditioning system, and improving the equipment life and user comfort.

[0103] Embodiment 2

[0104] A company is upgrading the technology of a household multi-split air conditioning system (one-to-four) and intends to adopt an air conditioning control system provided by the present invention, including:

[0105] A data acquisition module is used to collect the status parameters of each indoor unit in the multi-split system in real time; the status parameters include the inlet and outlet temperature and pressure of the evaporator, the inlet and outlet temperature and pressure of the condenser, the ambient temperature and humidity, and the current opening of the electronic expansion valve;

[0106] A refrigerant flow prediction module, used to calculate the required flow of each indoor unit based on the state parameters through a pre-trained refrigerant flow prediction model, and generate a global refrigerant allocation strategy in combination with the total system load;

[0107] A feedforward control module, configured to feedforward control the opening of the electronic expansion valve using a multivariable collaborative control algorithm according to the global refrigerant distribution strategy, and synchronously adjust the compressor frequency to match the system load with the flow rate;

[0108] The feedback control module is used to monitor the evaporator outlet superheat deviation and compressor frequency deviation, and control the electronic expansion valve opening and compressor frequency through PID controller feedback.

[0109] In order to evaluate the performance difference between the new and old systems, a performance test was carried out. The test environment was a high temperature condition in summer (outdoor temperature 35℃, indoor initial temperature 30℃, target temperature 25℃). The original air-conditioning control system was based on a traditional PID controller. The comparison of indicators between the original control system and the current control system is shown in Table 1. From the performance test results, the current control system is significantly better than the original control system in many key indicators. The energy efficiency ratio (EER) was improved by 28.1%, the temperature stabilization time was shortened by 52%, the superheat fluctuation range was greatly reduced by 77%, and the number of compressor starts and stops was reduced by 62.5%. These improvements show that the current control system not only improves energy efficiency and temperature control accuracy through the optimized global refrigerant distribution strategy and multivariable collaborative control algorithm, but also reduces system fluctuations and equipment losses, and further improves the stability and operation efficiency of the system. In summary, the current control system has significant advantages in reducing energy consumption, improving comfort, and extending equipment life.

[0110] Table 1 Comparison of indicators between the original control system and the current control system

[0111] index Original control system Current control system Energy Efficiency Ratio (EER) 3.2 4.1 Temperature stabilization time (minutes) 25 12 Superheat fluctuation range(℃) ±3.5 ±0.8 Compressor start and stop times / hour 8 3

[0112] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An air conditioning control method, characterized in that: include: Real-time collection of status parameters of each indoor unit in the multi-split system; the status parameters include evaporator inlet and outlet temperature and pressure, condenser inlet and outlet temperature and pressure, ambient temperature and humidity, and current opening of the electronic expansion valve; Based on the state parameters, the required flow of each indoor unit is calculated by a pre-trained refrigerant flow prediction model, and a global refrigerant allocation strategy is generated in combination with the total system load; According to the global refrigerant distribution strategy, a multivariable collaborative control algorithm is used to feedforward control the opening of the electronic expansion valve and synchronously adjust the compressor frequency to match the system load with the flow rate; Monitor the evaporator outlet superheat deviation and compressor frequency deviation, and control the electronic expansion valve opening and compressor frequency through PID controller feedback.

2. An air conditioning control method according to claim 1, characterized in that: The establishment of the refrigerant flow prediction model includes: Acquire historical data records of the indoor unit, and preprocess the historical data records to obtain preprocessed data records; the historical data records include historical state parameters and historical demand flow; The preprocessed data records are divided into a training set and a validation set in a ratio of 7:3, the training set is used to train the fusion model of LSTM combined with XGBoost, and the fusion model is optimized through a loss function; When the loss function converges, the fusion model is verified using a verification set; if the verification error is greater than an error threshold, the fusion model is retrained; otherwise, a refrigerant flow prediction model is obtained.

3. An air conditioning control method according to claim 2, characterized in that: The refrigerant flow prediction model includes: Input layer, used to receive the status parameters of the indoor unit; A fusion layer, including LSTM units and XGBoost units, is used to extract and integrate the time series features and static features of the state parameters; Prediction layer, used to output prediction results; The error feedback layer is used to execute the loss function and reduce the training error during model training; The output layer is used to output the required flow.

4. The air conditioning control method according to claim 1, characterized in that: The global refrigerant allocation strategy generated by combining the total system load includes: Calculate total system load and compressor target frequency; Establishing a global refrigerant allocation constraint function according to the real-time monitoring data and the compressor target frequency; the real-time monitoring data includes the set temperature, real-time temperature, real-time flow rate and required flow rate of the indoor unit; Solving the global refrigerant allocation constraint function to obtain a global refrigerant allocation strategy; The calculation formula of the global refrigerant allocation constraint function is: Among them, min represents the minimization function, i represents the index number of the indoor unit, T set,i represents the set temperature of the i-th indoor unit, T real,i represents the real-time temperature of the i-th indoor unit, m i represents the real-time flow of the i-th indoor unit, m pre,i represents the demand flow of the i-th indoor unit, f comp represents the compressor target frequency; ω1, ω2 and ω3 represent the temperature weight, flow weight and load weight, respectively.

5. The air conditioning control method according to claim 1, characterized in that: The feedforward control comprises: Defining system state variables and system control variables, and establishing state equations according to the system state variables and the system control variables; Solving the state equation to obtain the electronic expansion valve opening adjustment amount and the compressor frequency adjustment amount; The electronic expansion valve opening and the compressor frequency are adjusted according to the electronic expansion valve opening adjustment amount and the compressor frequency adjustment amount.

6. The air conditioning control method according to claim 1, characterized in that: The feedback control comprises: Calculate the superheat and superheat deviation at the evaporator outlet; Adjust the opening of the electronic expansion valve based on PID control to match the refrigerant flow to the load demand and ensure the stability of superheat; Adjust the compressor frequency based on the electronic expansion valve opening feedback to ensure that the refrigerant supply matches the system load demand and avoid frequent start and stop of the compressor; Continuously monitor system status and optimize gain coefficients.

7. An air conditioning control system, characterized in that: include: A data acquisition module is used to collect the status parameters of each indoor unit in the multi-split system in real time; the status parameters include the inlet and outlet temperature and pressure of the evaporator, the inlet and outlet temperature and pressure of the condenser, the ambient temperature and humidity, and the current opening of the electronic expansion valve; A refrigerant flow prediction module, used to calculate the required flow of each indoor unit based on the state parameters through a pre-trained refrigerant flow prediction model, and generate a global refrigerant allocation strategy in combination with the total system load; A feedforward control module, configured to feedforward control the opening of the electronic expansion valve using a multivariable collaborative control algorithm according to the global refrigerant distribution strategy, and synchronously adjust the compressor frequency to match the system load with the flow rate; The feedback control module is used to monitor the evaporator outlet superheat deviation and compressor frequency deviation, and control the electronic expansion valve opening and compressor frequency through PID controller feedback.

Images