Flexible load control method and system for multi-split variable frequency air conditioner and medium

Abstract

The invention discloses a flexible load control method and system for a multi-split variable frequency air conditioner and a medium. The method comprises the steps that operation data are collected; building an air conditioner power model: building the air conditioner power model based on the collected operation data; and air conditioner load response: when the power instruction is received, the air conditioner load response is started, and the flexible load control strategy is executed on the multi-split variable frequency air conditioner. According to the method, load control suitable for VRV air conditioners of different types and powers is achieved, the method is suitable for rooms of different state conditions, and the later maintenance cost is saved.

Description

Technical Field

[0001] The present invention relates to the technical field of building energy management, and in particular to a flexible load control method, system and medium for a multi-split variable frequency air conditioner based on a power model. Background Art

[0002] Multi-split inverter air conditioner (VRV) is a central air conditioner with one outdoor unit connected to multiple indoor units. It has the advantages of easy installation, small footprint, wide applicable temperature range, separate control of different indoor units, stable and comfortable operation, etc. It is widely distributed in small and medium-sized hotels, guesthouses, office buildings, hospitals, villas and other occasions. VRV is an important part of the building load, and the control of its load is one of the important methods to achieve building energy conservation and carbon reduction, peak shifting and fee reduction.

[0003] VRV air conditioner flexible load regulation refers to responding to grid load regulation instructions without completely shutting down the air conditioner load. Because VRV air conditioners cannot directly control power, they need to indirectly control air conditioner power by adjusting the set temperature, duty cycle, and peak start and stop. The setting of the adjustment temperature is limited to the human body's comfortable temperature range, which is generally between 23.5-28.5℃, so the adjustable range of the air conditioner flexible load is limited.

[0004] The flexible load regulation of VRV air conditioners is limited by two problems:

[0005] 1. The flexible load adjustment range is unclear, and it is difficult to determine the set temperature to be selected. Because the flexible load adjustment range is not only a function of the set temperature, but also related to the actual indoor temperature, outdoor temperature, room heat storage, indoor heat load real-time power, personnel activity disturbance, etc. Some variables are difficult to measure or quantify, and they change with the use status of the building.

[0006] 2. Power rebound during the regulation process. Because the air conditioner set temperature interval is usually between 0.5-1℃, it is difficult to find a temperature that balances the cooling load. As the cooling load decreases after load regulation, the heat stored in various items in the room will be gradually released, causing the room temperature to rise and the cooling load demand to increase again. This makes the load response stability of the air conditioner insufficient. Summary of the invention

[0007] The main purpose of the present invention is to propose a flexible load control method, system and medium for a multi-split variable frequency air conditioner, aiming to realize load control of VRV air conditioners of different models and powers, suitable for rooms with different state conditions, and save subsequent maintenance costs.

[0008] To achieve the above object, the present invention provides a flexible load control method for a multi-split variable frequency air conditioner, the method comprising the following steps:

[0009] Step 1, operation data collection: obtain the real-time total power P of the air conditioner or the outdoor unit power, and the actual indoor temperature Indoor unit set temperature Actual outdoor temperature T out , weather type weather, set temperature change amount and change time tk and corresponding indoor unit number, shutdown number and shutdown time of some indoor units, continuous monitoring data;

[0010] Among them, the superscript N represents the indoor unit number; the subscript in represents the actual indoor temperature of the room, the subscript set represents the set temperature of the indoor unit of the room, and the subscript out represents the actual outdoor temperature. They are used to distinguish the nature of the temperature; tk is the time mark according to the 24-hour system;

[0011] Step 2, air conditioning power model construction: build the air conditioning power model based on the collected operation data;

[0012] Step 3, air conditioning load response: when receiving the power instruction, start the air conditioning load response and execute the flexible load control strategy for the multi-split variable frequency air conditioner, wherein the power instruction includes the allowable power upper limit and power control time, the configuration parameters include the comfort temperature range, the air conditioner settable temperature vector, and the operating parameters include the indoor temperature, outdoor temperature, and time information.

[0013] A further technical solution of the present invention is that in the step Step 2, the data set is split according to weather type, different weather types are trained for different power types, and each time the power model is called, the corresponding weather type must be selected, and the step Step 2 includes:

[0014] Step S210, power signal filtering processing: filtering the actual power based on the sliding average filtering method, wherein the filtering sliding time is the multi-split variable frequency air conditioner power up and down adjustment cycle, and the sliding average filtering formula of the actual power of the air conditioner is:

[0015]

[0016] Among them, Δt represents the power measurement interval, and n represents the power adjustment cycle of the multi-split variable frequency air conditioner;

[0017] Step S220, temperature signal frame interpolation processing: interpolate the actual temperature frame based on the difference method to ensure that each power has a corresponding actual value of the indoor and outdoor temperature, wherein the temperature signal frame interpolation formula is:

[0018]

[0019] Among them, t0 and t1 are two moments with temperature measurements, the subscript "interpolatory" indicates that the temperature is obtained by interpolation, and tk is the moment between t0 and t1 that needs to be interpolated.

[0020] Step 8230, power modeling is performed based on artificial neural network: the data set is split according to weather type, and different power models are trained for different weather types; the input vectors are the actual indoor temperature, the set temperature of the indoor unit, the actual outdoor temperature, and the time; the output variable is the average power after normalization, and the normalization benchmark is the maximum working power of the air conditioner; because the air conditioner power is nonlinear within a limited range, the Sigmoid function is selected as the activation function;

[0021] Among them, the input and output formula of the air conditioning power model is:

[0022]

[0023] The activation function formula of artificial neural network is:

[0024]

[0025] The subscript "rated" indicates the maximum rated value. It is the normalization of power. f refers to the generalized undetermined function. The subscript ANN indicates that the undetermined function is constructed based on a neural network. [] indicates that the input is a set of vectors. The subscript tk indicates that all variables are quantities corresponding to time tk. S(x) is the component function form of the hidden layer in ANN. The right side is the standard expression of Sigmoid, where exp represents the exponential function of natural constants, b is the undetermined constant, and a1 to a n Represents the unknown coefficients, x1 to x n Refers to the variables in [] in the input and output formula of the air conditioning power model.

[0026] A further technical solution of the present invention is that the step 3 comprises:

[0027] Step S310, judging: based on the air-conditioning power model, under the current indoor parameter state, predict the power range within the comfortable adjustable temperature range. If the power command is within the power range, it is judged that it can be responded. If the power command is outside the power range, it is inquired whether the response residual is allowed. If it is allowed, it responds according to the boundary value of the air-conditioning comfortable temperature range and feeds back the response residual. Otherwise, the response execution is exited. In the cooling mode, the load response judgment formula is:

[0028]

[0029] Among them, the subscript comfort represents the comfort range; up_limit represents the upper limit of the comfort temperature range; dn_limit represents the lower limit of the comfort temperature range; the subscript command represents the command, P command Indicates power command;

[0030] Step 8320, selection: based on the air conditioning power model, select the set temperature that is closest to the power command, wherein all settable temperatures are traversed, and the value with the predicted power less than the power command and the smallest load response error is selected, and the partial shutdown strategy is not considered in the initial response. The pseudo code for selecting the set temperature is:

[0031]

[0032] Among them, || indicates absolute value calculation, min indicates taking the smaller value of two calculated values, and Error indicates the response residual; Lists_T set Represents a set of all optional set temperatures for a multi-split variable-frequency air conditioner, configured according to the air conditioner hardware attributes; Ti is an intermediate variable for temperature calculation, T is the selected set temperature value, and return is a command to return to the program;

[0033] Step S330, execution: first send the set temperature to each indoor unit; and continuously monitor the actual power of the air conditioner within the power command control time range, and calculate the sliding average power; the end condition of the execution must meet any one of the following conditions: 1. The power control time ends; 2. The indoor temperature exceeds the comfort temperature range limit for the first preset time period;

[0034] Step S340, tracking feedback: suppress power rebound through tracking feedback, and the criterion for tracking feedback is: starting from the second preset time after the power instruction is issued, if the sliding average power value of the preset number of times in the past third preset time is greater than the power instruction, the suppression strategy is triggered; the set temperature is calculated again according to the air-conditioning power model and the pseudo code for selecting the set temperature; if the set temperature has a value, the set temperature adjustment instruction is issued according to the new set temperature; if the set temperature is empty, an instruction to shut down some indoor units is issued, and the indoor units with low air inlet temperatures are shut down one by one; until the power rebound is suppressed, or the indoor temperature exceeds the comfortable temperature range.

[0035] To achieve the above object, the present invention further proposes a multi-split variable frequency air conditioner flexible load control system, the system comprising a memory, a processor, and a multi-split variable frequency air conditioner flexible load control program stored on the processor, the multi-split variable frequency air conditioner flexible load control program being executed by the processor to perform the following steps:

[0036] Step 1, operation data collection: obtain the real-time total power P of the air conditioner or the outdoor unit power, and the actual indoor temperature Indoor unit set temperature Actual outdoor temperature T out , weather type weather, set temperature change amount and change time tk and corresponding indoor unit number, shutdown number and shutdown time of some indoor units, continuous monitoring data;

[0037] Among them, the superscript N represents the indoor unit number; the subscript in represents the actual indoor temperature of the room, the subscript set represents the set temperature of the indoor unit of the room, and the subscript out represents the actual outdoor temperature. They are used to distinguish the nature of the temperature; tk is the time mark according to the 24-hour system;

[0038] Step 2, air conditioning power model construction: build the air conditioning power model based on the collected operation data;

[0039] Step 3, air conditioning load response: when receiving the power instruction, start the air conditioning load response and execute the flexible load control strategy for the multi-split variable frequency air conditioner, wherein the power instruction includes the allowable power upper limit and power control time, the configuration parameters include the comfort temperature range, the air conditioner settable temperature vector, and the operating parameters include the indoor temperature, outdoor temperature, and time information.

[0040] A further technical solution of the present invention is that when the multi-split variable frequency air conditioner flexible load control program is executed by the processor, the following steps are further performed:

[0041] Step S210, power signal filtering processing: filtering the actual power based on the sliding average filtering method, wherein the filtering sliding time is the multi-split variable frequency air conditioner power up and down adjustment cycle, and the sliding average filtering formula of the actual power of the air conditioner is:

[0042]

[0043] Among them, Δt represents the power measurement interval, and n represents the power adjustment cycle of the multi-split variable frequency air conditioner;

[0044] Step S220, temperature signal frame interpolation processing: interpolate the actual temperature frame based on the difference method to ensure that each power has a corresponding actual value of the indoor and outdoor temperature, wherein the temperature signal frame interpolation formula is:

[0045]

[0046] Among them, t0 and t1 are two moments with temperature measurements, the subscript "interpolatory" indicates that the temperature is obtained by interpolation, and tk is the moment between t0 and t1 that needs to be interpolated.

[0047] Step S230, power modeling is performed based on artificial neural network: the data set is split according to weather type, and different power models are trained for different weather types; the input vectors are the actual indoor temperature, the set temperature of the indoor unit, the actual outdoor temperature, and the time; the output variable is the average power after normalization, and the normalization benchmark is the maximum working power of the air conditioner; since the air conditioner power is nonlinear within a limited range, the Sigmoid function is selected as the activation function;

[0048] Among them, the input and output formula of the air conditioning power model is:

[0049]

[0050] The activation function formula of artificial neural network is:

[0051]

[0052] The subscript "rated" indicates the maximum rated value. It is a standardized processing of power. f refers to a generalized undetermined function. The subscript ANN indicates that this undetermined function is constructed based on a neural network. [] indicates that the input quantity is a set of vectors. The subscript tk indicates that all variables are quantities corresponding to time tk. S(x) is the component function form of the hidden layer in ANN. is the standard expression of Sigmoid, where exp represents the exponential function of a natural constant, b is an undetermined constant, and a1 to a n Represents the unknown coefficients, x1 to x n Refers to the variables in [] in the input and output formula of the air conditioning power model.

[0053] A further technical solution of the present invention is that when the multi-split variable frequency air conditioner flexible load control program is executed by the processor, the following steps are performed:

[0054] Step S310, judging: based on the air-conditioning power model, under the current indoor parameter state, predict the power range within the comfortable adjustable temperature range. If the power command is within the power range, it is judged that it can be responded. If the power command is outside the power range, it is inquired whether the response residual is allowed. If it is allowed, it responds according to the boundary value of the air-conditioning comfortable temperature range and feeds back the response residual. Otherwise, the response execution is exited. In the cooling mode, the load response judgment formula is:

[0055]

[0056] Among them, the subscript comfort represents the comfort range; up_limit represents the upper limit of the comfort temperature range; dn_limit represents the lower limit of the comfort temperature range; the subscript command represents the command, P command Indicates power command;

[0057] Step S320, selection: based on the air conditioning power model, select the set temperature that is closest to the power command, wherein all settable temperatures are traversed, and the value with the predicted power less than the power command and the smallest load response error is selected, and the partial shutdown strategy is not considered in the initial response. The pseudo code for selecting the set temperature is:

[0058]

[0059] Among them, || indicates absolute value calculation, min indicates taking the smaller value of two calculated values, and Error indicates the response residual; Lists_T set Represents a set of all optional set temperatures for a multi-split variable-frequency air conditioner, configured according to the air conditioner hardware attributes; Ti is an intermediate variable for temperature calculation, T is the selected set temperature value, and return is a command to return to the program;

[0060] Step S330, execution: first send the set temperature to each indoor unit; and continuously monitor the actual power of the air conditioner within the power command control time range, and calculate the sliding average power; the end condition of the execution must meet any one of the following conditions: 1. The power control time ends; 2. The indoor temperature exceeds the comfort temperature range limit for the first preset time period;

[0061] Step S340, tracking feedback: suppress power rebound through tracking feedback, and the criterion for tracking feedback is: starting from the second preset time after the power instruction is issued, if the sliding average power value of the preset number of times in the past third preset time is greater than the power instruction, the suppression strategy is triggered; the set temperature is calculated again according to the air-conditioning power model and the pseudo code for selecting the set temperature; if the set temperature has a value, the set temperature adjustment instruction is issued according to the new set temperature; if the set temperature is empty, an instruction to shut down some indoor units is issued, and the indoor units with low air inlet temperatures are shut down one by one; until the power rebound is suppressed, or the indoor temperature exceeds the comfortable temperature range.

[0062] To achieve the above objectives, the present invention also proposes a computer-readable storage medium, which stores a multi-split variable frequency air conditioner flexible load control program, and when the multi-split variable frequency air conditioner flexible load control program is run by a processor, the steps of the method described above are executed.

[0063] The beneficial effects of the flexible load control method, system and medium of the multi-split variable frequency air conditioner of the present invention are:

[0064] 1. The algorithm has good adaptability and can adapt to VRV air conditioners of different models and powers and room environments with different thermal capacity and thermal resistance properties; it can automatically adapt to the zero drift of the air conditioner's cooling capacity and changes in the room status, reducing the cost of later debugging and maintenance.

[0065] 2. Minimize the method of directly shutting down the air conditioner, make full use of the flexible load under the set temperature change, and achieve a balance between load response and temperature comfort assurance.

[0066] 3. An innovative method for suppressing air conditioning power rebound is discussed. Based on continuous temperature adjustment and partial shutdown of indoor units, a new relatively balanced cooling load state can be gradually found based on this method, despite the interference of difficulties such as strong thermal inertia and strong heat storage interference in the room. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0068] Figure 1 It is a flow chart of the flexible load control method of the multi-split variable frequency air conditioner of the present invention;

[0069] Figure 2 This is the step composition diagram of step 1 running data collection;

[0070] Figure 3 It is the step composition diagram of the power model construction in step Step2;

[0071] Figure 4 This is the measured result of signal filtering in Step 2;

[0072] Figure 5 It is the function structure diagram of the power model based on ANN in step 2;

[0073] Figure 6 It is the measured result diagram of the power prediction value under the variable working condition of the power model in step 2;

[0074] Figure 7 It is the step composition diagram of the air conditioning load response in step Step3;

[0075] Figure 8 This is the measured result diagram of the power rebound suppression strategy 1 in Step 3 based on the continuous set temperature adjustment;

[0076] Fig. 9 It is the measured result diagram of power rebound suppression strategy 2 in step 3 based on partially shutting down the indoor units;

[0077] Fig.10It is a schematic diagram of a VRV flexible load controller that executes power model construction and flexible load control algorithm.

[0078] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0079] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0080] In order to realize load control of VRV air conditioners of different models and powers, suitable for rooms with different state conditions, and save later maintenance costs, the present invention proposes a flexible load control method for multi-split variable frequency air conditioners. The technical scheme adopted by the flexible load control method for multi-split variable frequency air conditioners proposed by the present invention is: based on the measuring points of the air conditioner flexible load control module, the operating parameters are tracked and recorded; the power model is modeled by an artificial neural network (ANN) with strong nonlinear adaptability and good fault tolerance; in the flexible load response control process, the change of the set temperature is used as much as possible to respond to the power instruction, and the balance between power and indoor temperature comfort is achieved as much as possible without completely shutting down the VRV air conditioner; and two methods of continuous set temperature adjustment and partial shutdown of the air conditioner indoor unit are proposed to suppress power rebound.

[0081] To achieve the above objectives, the present invention specifically provides a set of technical solutions covering the arrangement of measuring point systems, data processing and data modeling, detailed implementation process of air conditioning load response and related control device media.

[0082] Please refer to Figures 1 to 10 The preferred embodiment of the multi-split variable frequency air conditioner flexible load control method of the present invention comprises the following steps:

[0083] Step 1, operation data collection: obtain the real-time total power P of the air conditioner or the outdoor unit power, and the actual indoor temperature Indoor unit set temperature T set , Actual outdoor temperature T out , weather type weather, set temperature change amount and change time tk and corresponding indoor unit number, shutdown number and shutdown time of some indoor units, continuous monitoring data.

[0084] Among them, the superscript N represents the indoor unit number; the subscript in represents the actual indoor temperature of the room, the subscript set represents the set temperature of the room indoor unit, and the subscript out represents the actual outdoor temperature. They are used to distinguish the nature of the temperature; tk is the time mark according to the 24-hour system.

[0085] In this embodiment, the operation data can be collected through the measuring point system, as follows:

[0086] Air conditioning power measurement point; VRV total power measurement point, recording the real-time total power of the air conditioner; because the indoor unit is mainly blowing power, the working power is very low and there is no power adjustable range. If the indoor unit and the outdoor unit are not the same power source, only the outdoor unit power can be measured; the power meter requires an accuracy higher than 0.01kW, a measurement frequency greater than or equal to 1 time / min, and good telemetry communication.

[0087] Temperature measuring points: indoor temperature measuring points, one measuring point is placed at the temperature air inlet of each air-conditioning indoor unit; outdoor temperature measuring points are arranged in an outdoor louvered box, the louvered box should be well ventilated, unobstructed and out of direct sunlight, and the height from the ground should be greater than 1 meter; the temperature meter is required to have an accuracy higher than 0.1℃, a measurement frequency greater than or equal to 1 time / 5min, and good telemetry communication.

[0088] Record the set temperature; record the set temperature of each air-conditioning indoor unit. VRV recommends synchronous changes in the set temperature of multiple indoor units; record the change amount and change time of the set temperature and the corresponding indoor unit number; record the shutdown number and shutdown time of some indoor units, with the time accurate to the minute.

[0089] Record weather conditions as the basis for data grouping; can be based on weather information service providers or solar radiation, rainfall sensors for recording, mainly used to qualitatively distinguish between sunny or cloudy, overcast, rainy or snowy weather types, accurate to 15 minutes.

[0090] Continuous monitoring data: When the number of power measurement records for a certain weather type exceeds 2400, that is, more than 40 hours, the data of this weather type can enter Step_2; the variable labels of the continuous monitoring data are as follows:

[0091] P Air Conditioning Real-time Power

[0092] Actual indoor temperature (N is the indoor unit number)

[0093] Indoor unit set temperature (N is the indoor unit number)

[0094] T out Actual outdoor temperature

[0095] weather weather type

[0096] tk=hh:mm time (accurate to minute)

[0097] Among them, the superscript N represents the indoor unit number (1, 2, 3), because VRV air conditioners have multiple indoor units and need to be numbered; the subscript in represents the actual indoor temperature of the room, the subscript set represents the set temperature of the room indoor unit, and the subscript out represents the actual outdoor temperature. They are used to distinguish the nature of the temperature; tk is the time mark according to the 24-hour system, for example, 2:30 in the afternoon is marked as tk=14:30, where hh=14 and mm=30.

[0098] Step 2, air conditioning power model construction: build the air conditioning power model based on the collected operation data.

[0099] In the step 2, the data set is split according to weather types, and different power types are trained for different weather types. Each time the power model is called, the corresponding weather type must be selected.

[0100] The step 2 comprises:

[0101] Step S210, power signal filtering processing: filtering the actual power based on the sliding average filtering method, wherein the filtering sliding time is the power up and down adjustment cycle of the multi-split variable frequency air conditioner.

[0102] The power signal filtering process is similar to the "duty cycle" periodic start and stop of the compressor of a fixed-frequency air conditioner. The power of a VRV variable-frequency air conditioner is not completely steplessly adjustable. It has multiple power gears and periodically adjusts the power up and down within a certain power range based on the power gear, but the corresponding power gear does not directly shut down the compressor; this causes additional errors due to the periodicity of the power, and the actual power is filtered based on the sliding average filtering method; the VRV power up and down adjustment cycle is usually within 10 minutes, so the filtering sliding time is 10 minutes. Although the increase in the sliding time makes the filtering effect better, it will cause the instantaneous accuracy of the power model to decrease.

[0103] The sliding average filter formula of the actual power of the air conditioner is:

[0104]

[0105] Among them, Δt represents the power measurement interval, and n represents the power adjustment cycle of the multi-split variable frequency air conditioner.

[0106] Usually Δt=1min, n=10; for example, when tk=14:30, n=10, It indicates the average power of the 10 minutes from 14:21 to 14:30.

[0107] Step S220, temperature signal frame interpolation processing: interpolate the actual temperature frame based on the difference method to ensure that each power has a corresponding actual value of the indoor and outdoor temperature.

[0108] Indoor thermal inertia is generally high, and the temperature changes slower than the power, so the temperature measurement frequency is lower than the power measurement frequency, which can reduce communication pressure and data processing pressure. To ensure that each power has a corresponding actual value of the indoor and outdoor temperature, the actual temperature is interpolated based on the difference method.

[0109] Among them, the temperature signal frame filling formula is:

[0110]

[0111] Among them, t0 and t1 are two moments with temperature measurement values, the subscript "interpolatory" indicates that the temperature is obtained by interpolation, and tk is the moment between t0 and t1 that needs to be interpolated.

[0112] For example, at t0=14:30, the temperature T is measured. t0 =23.5℃, t1=14:35, measured temperature T t1 =24℃, according to the formula tk=14:32 when the frame temperature

[0113] Step 8230, power model building is performed based on artificial neural network: the data set is split by weather type, and different power models are trained for different weather types; the input vectors are the actual indoor temperature, the set temperature of the indoor unit, the actual outdoor temperature, and the time; the output variable is the average power after normalization, and the normalization benchmark is the maximum working power of the air conditioner; because the air conditioner power is nonlinear within a limited range, the Sigmoid function is selected as the activation function.

[0114] Among them, the input and output formula of the air conditioning power model is:

[0115]

[0116] The activation function formula of artificial neural network is:

[0117]

[0118] The subscript "rated" indicates the maximum rated value. It is a standardized processing of power. f refers to a generalized undetermined function. The subscript ANN indicates that this undetermined function is constructed based on a neural network. [] indicates that the input quantity is a set of vectors. The subscript tk indicates that all variables are quantities corresponding to time tk. S(x) is the component function form of the hidden layer in ANN. is the standard expression of Sigmoid, where exp represents the exponential function of a natural constant, b is an undetermined constant, and a1 to a n Represents the unknown coefficients, x1 to x n Refers to the variables in [] in the input and output formula of the air conditioning power model; the meanings of other quantities such as N, in, out, hh, etc. are the same as those in the previous text.

[0119] The trained artificial neural network is used to predict the air-conditioning power under different environmental conditions and different set temperatures. As the operating data and load response logs are accumulated, the model needs to be updated regularly to continuously adapt to changes in the life of the air-conditioning equipment and the indoor and outdoor environment of the room.

[0120] Step 3, air conditioning load response: when receiving the power instruction, start the air conditioning load response and execute the flexible load control strategy for the multi-split variable frequency air conditioner, wherein the power instruction includes the allowable power upper limit and power control time, the configuration parameters include the comfort temperature range, the air conditioner settable temperature vector, and the operating parameters include the indoor temperature, outdoor temperature, and time information.

[0121] The main principle of the VRV air conditioner flexible load response control algorithm is to use the change of the set temperature to respond to the power command as much as possible, and not to completely shut down the VRV air conditioner; to achieve a balance between power and indoor temperature comfort as much as possible.

[0122] In this embodiment, the starting condition of the load response is receiving the power command. After receiving the power dispatch command, the air conditioning load response is divided into four steps: judgment, selection, execution, and tracking feedback. The main basis for judgment is the load response judgment formula, and the response residual is predicted; the main basis for selection is to select the set temperature; the execution process mainly tracks the time of the load response command and the indoor temperature comfort, and judges the execution end condition; the tracking feedback is mainly to ensure that the power rebound does not exceed the limit.

[0123] Among them, the air-conditioning power model to be constructed is required to be deeply involved in the steps of judgment, selection, tracking and feedback; in the judgment, the power range corresponding to the upper and lower limits of the comfort temperature range determines whether the air-conditioning participates in load response and the participation limit; in the selection and tracking feedback, it is the basis for determining the set temperature.

[0124] The step 3 specifically includes:

[0125] Step S310, judging: based on the air-conditioning power model, under the current indoor parameter state, predict the power range within the comfortable adjustable temperature range. If the power command is within the power range, it is judged that it can be responded. If the power command is outside the power range, it is inquired whether the response residual is allowed. If it is allowed, it responds according to the boundary value of the air-conditioning comfortable temperature range and feeds back the response residual. Otherwise, the response execution is exited. In the cooling mode, the load response judgment formula is:

[0126]

[0127] Among them, the subscript comfort represents the comfort range; up_limit represents the upper limit of the comfortable temperature range. Generally, T set =28.5℃ or 29℃, which is configured by the air conditioner user; dn_limit indicates the lower limit of the comfortable temperature range. Generally, T set =23.5℃ or 23℃, which is configured by the air conditioner user; the superscript command indicates the command, P command Indicates power command;

[0128] Step S320, selection: based on the air conditioning power model, select the set temperature that is closest to the power instruction, wherein all settable temperatures are traversed, and the value with the predicted power less than the power instruction and the smallest load response error is selected, and the partial shutdown strategy is not considered in the initial response.

[0129] The selection process is also based on the power model, and the set temperature that is closest to the power command is selected; the partial shutdown strategy is not considered during the initial response; because the air conditioner set temperature interval is 0.5-1℃, it is only necessary to traverse all the settable temperatures and select the value with the predicted power less than the power command and the smallest load response error.

[0130] The pseudo code for selecting the set temperature is:

[0131]

[0132] Among them, || indicates absolute value calculation, min indicates taking the smaller value of two calculated values, and Error indicates the response residual; Lists_T set Represents the set of all optional set temperatures for multi-split inverter air conditioners, configured according to the air conditioner hardware properties; Ti is the intermediate variable for temperature calculation, T is the selected set temperature value, and return is the return program command. The purpose of this algorithm is to select the set temperature value with the smallest response residual when the actual power is less than the power command.

[0133] Step S330, execution: first send the set temperature to each indoor unit; and continuously monitor the actual power of the air conditioner within the power command control time range, and calculate the sliding average power; the end condition of the execution must meet any one of the following conditions: 1. The power control time ends; 2. The indoor temperature exceeds the comfort temperature range limit for the first preset time.

[0134] Among them, in the second condition, the first preset time is set to 10 minutes. If the indoor temperature exceeds the comfortable temperature range limit for 10 consecutive minutes, the execution is terminated. If any of the above two conditions is met, the power response is terminated and the current set temperature state remains unchanged.

[0135] Step S340, tracking feedback: suppress power rebound through tracking feedback, and the criterion for tracking feedback is: starting from the second preset time after the power instruction is issued, if the sliding average power value of the preset number of times in the past third preset time is greater than the power instruction, the suppression strategy is triggered; the set temperature is calculated again according to the air-conditioning power model and the pseudo code for selecting the set temperature; if the set temperature has a value, the set temperature adjustment instruction is issued according to the new set temperature; if the set temperature is empty, an instruction to shut down some indoor units is issued, and the indoor units with low air inlet temperatures are shut down one by one; until the power rebound is suppressed, or the indoor temperature exceeds the comfortable temperature range.

[0136] The second preset time can be set to 6 minutes, the third preset time can be set to 5 minutes, and the cumulative preset number is set to 4 times. Starting from the 6th minute after the power command is issued, if the cumulative sliding average power values ​​of the past 5 minutes for 4 times are all greater than the power command, the suppression strategy is triggered.

[0137] The tracking feedback in this embodiment is mainly to suppress power rebound, and the execution strategies include: 1. Continuously changing the set temperature in small amplitudes; 2. Partially shutting down the indoor unit.

[0138] The tracking feedback is based on the air conditioning model calculation, calculates the new set temperature, and determines which execution strategy to select; if there is a new set temperature value, then the new set temperature is followed, that is, the set temperature strategy is changed continuously and slightly; if the set temperature is empty, the indoor units with low air inlet temperatures are closed one by one, that is, the strategy of partially closing the indoor units is adopted.

[0139] The termination condition of the tracking feedback to suppress power rebound is: until the power rebound is suppressed, or the indoor temperature exceeds the comfortable temperature range.

[0140] In this embodiment, step Step 1 is the data basis of step Step 2 and step Step 3, and step Step 2 is the calculation basis of step Step 3; step Step 1 includes a specific measurement point layout system and requirements; the collected real-time data provides input quantities and controlled target quantities for step Step 3 to execute flexible load response; the real-time data is regularly summarized as an operation log to provide a training set for the construction of the power model of step Step 2; step Step 2 is the basis for the function calculation in each step of step Step 3.

[0141] In this embodiment, the calculation and execution of step Step 2 and step Step 3 are encapsulated in the same control medium, the VRV flexible load controller; specifically, it includes at least four parts, namely, an operation log storage, a power model calculator, a power command executor, and a timer; the operation log storage module receives various data from the measurement point system, and the power command executor receives the power command issued; the power command executor issues a set temperature adjustment or partial indoor unit shutdown command to the VRV air conditioner through infrared communication or a power control panel.

[0142] The following combination Figures 1 to 10 The flexible load control method of the multi-split variable frequency air conditioner of the present invention is further explained.

[0143] See also Figure 1 The implementation case of the present invention includes three steps: operation data collection, power model construction, and air conditioning load response. A three-to-one VRV variable frequency air conditioner with a maximum power of 16.89kW is selected for empirical testing of the present invention.

[0144] See also Figure 2 The specific steps for collecting operation data are as follows: power meters are installed on the indoor and outdoor units respectively, the measurement frequency of the power meter is 1 time / min, and the measurement accuracy is 0.01kW; a temperature probe is installed at the air inlet of each indoor unit, a louvered box is arranged outdoors, and a temperature probe is arranged in it. The measurement frequency of each temperature probe is 1 time / 5min, and the measurement accuracy is 0.001℃; 5 sunny days in summer are selected for continuous measurement to ensure uniform weather types; the air conditioner runs for 8 to 10 hours every day to ensure the data volume.

[0145] See also Figure 3 ,The power model construction includes data processing and data modeling; data processing mainly includes signal frame filling and signal noise removal, and preparing the training set data and target set data of the model; data modeling is based on artificial neural network modeling, training the training set and target set to establish the functional relationship between the set temperature and the power under different states of the room.

[0146] See also Figure 4, respectively showing the measured power values ​​and the power values ​​after sliding average filtering; VRV air conditioners include at least 2.3kW, 5.5kW, 11.7kW, 15.5kW, 16.5kW and other power levels, and the power fluctuates periodically; with 10min as the sliding window time and the sliding average filtering formula, it can be seen that the power fluctuation error is greatly reduced, and the relative deviation drops from more than 15% to less than 10%.

[0147] Regarding the signal frame filling algorithm, since the temperature measurement value is one point every 5 minutes, there are 4 points in the 5 power points within 5 minutes without corresponding temperatures, so it is necessary to use interpolation method between every two temperature measurement points to fill in the other 4 missing temperatures; the method is based on the temperature frame filling formula, and calculates according to the relationship between the time of the framed temperature and the time between the two adjacent temperature measurement points to achieve a one-to-one correspondence between the power measurement points and the temperature measurement points.

[0148] See also Figure 5 , showing the internal structure of the artificial neural network and the composition of the training set and output set; the artificial neural network includes an input layer, a hidden layer, and an output layer; the activation function formula of the hidden layer is the Sigmoid formula, and the trained model is shown in the input and output formula of the air conditioning power model.

[0149] See also Figure 6 , showing the training results of the artificial neural network; realizing the prediction of the relationship between the set temperature and the air conditioning power; conversely, under the condition of determining certain indoor and outdoor temperature, time and other state parameters, based on the graph, the set temperature value closest to a certain power can be determined; the set temperature interval of the graph is 1°C, for air conditioners with a set temperature interval of 0.5°C, the artificial neural network can give a graph with a denser set temperature interval.

[0150] See also Figure 7 , shows a detailed flow chart of the air conditioning flexible load response control algorithm, which specifically includes four steps: judgment, selection, execution, and tracking feedback; the main basis for judgment is the load response judgment formula, and the response residual is predicted; the main basis for selection is Figure 6 The pseudo code of the model prediction value and the selection of the set temperature; the execution process mainly tracks the time of the load response instruction and the indoor temperature comfort; when the load response instruction time ends, the load response ends on time; when the indoor temperature comfort exceeds the limit, the load response ends in advance and the response residual is fed back; the tracking feedback mainly tracks the power rebound to ensure that the power rebound does not exceed the limit.

[0151] See also Figure 8 , Fig. 9 ,Two methods to effectively suppress the power rebound are demonstrated; Figure 8 The method of gradually finding the set temperature under the new heat exchange equilibrium state by continuously changing the set temperature in small increments was demonstrated, which gradually suppressed the power rebound. Fig. 9 It demonstrates the rapid power reduction achieved by partially shutting down indoor units and quickly adjusting the power down to meet the response demand, but this method may lead to a rapid increase in indoor temperature, affecting indoor temperature comfort and triggering load response interruption.

[0152] See also Fig.10 , showing the deployment methods of various parameter measurement points and VRV flexible load controllers; when deploying the algorithm, edge computing controllers can be arranged on site, and outdoor temperature measurement points, power meters, and air-conditioning equipment are all ready-made, minimizing the difficulty of construction; the air-conditioning control panel is similar to a fixed air-conditioning remote control, and uses the same communication interface and communication protocol as the air-conditioning remote control; in this way, the implementation of the VRV flexible load control algorithm can be realized on the spot.

[0153] The beneficial effects of the flexible load control method of the multi-split frequency conversion air conditioner of the present invention are:

[0154] 1. The algorithm has good adaptability and can adapt to VRV air conditioners of different models and powers and room environments with different thermal capacity and thermal resistance properties; it can automatically adapt to the zero drift of the air conditioner's cooling capacity and changes in the room status, reducing the cost of later debugging and maintenance.

[0155] 2. Minimize the method of directly shutting down the air conditioner, make full use of the flexible load under the set temperature change, and achieve a balance between load response and temperature comfort assurance.

[0156] 3. An innovative method for suppressing air conditioning power rebound is discussed. Based on continuous temperature adjustment and partial shutdown of indoor units, a new relatively balanced cooling load state can be gradually found based on this method, despite the interference of difficulties such as strong thermal inertia and strong heat storage interference in the room.

[0157] To achieve the above object, the present invention further proposes a multi-split variable frequency air conditioner flexible load control system, the system comprising a memory, a processor, and a multi-split variable frequency air conditioner flexible load control program stored on the processor, the multi-split variable frequency air conditioner flexible load control program being executed by the processor to perform the following steps:

[0158] Step 1, operation data collection: obtain the real-time total power P of the air conditioner or the outdoor unit power, and the actual indoor temperature Indoor unit set temperature Actual outdoor temperature T out , weather type weather, set temperature change amount and change time tk and corresponding indoor unit number, shutdown number and shutdown time of some indoor units, continuous monitoring data;

[0159] Among them, the superscript N represents the indoor unit number; the subscript in represents the actual indoor temperature of the room, the subscript set represents the set temperature of the indoor unit of the room, and the subscript out represents the actual outdoor temperature. They are used to distinguish the nature of the temperature; tk is the time mark according to the 24-hour system;

[0160] Step 2, air conditioning power model construction: build the air conditioning power model based on the collected operation data;

[0161] Step 3, air conditioning load response: when receiving the power instruction, start the air conditioning load response and execute the flexible load control strategy for the multi-split variable frequency air conditioner, wherein the power instruction includes the allowable power upper limit and power control time, the configuration parameters include the comfort temperature range, the air conditioner settable temperature vector, and the operating parameters include the indoor temperature, outdoor temperature, and time information.

[0162] Furthermore, when the multi-split variable frequency air conditioner flexible load control program is executed by the processor, the following steps are also performed:

[0163] Step S210, power signal filtering processing: filtering the actual power based on the sliding average filtering method, wherein the filtering sliding time is the multi-split variable frequency air conditioner power up and down adjustment cycle, and the sliding average filtering formula of the actual power of the air conditioner is:

[0164]

[0165] Among them, Δt represents the power measurement interval, and n represents the power adjustment cycle of the multi-split variable frequency air conditioner;

[0166] Step 8220, temperature signal frame interpolation processing: interpolate the actual temperature frame based on the difference method to ensure that each power has a corresponding actual value of the indoor and outdoor temperature. The temperature signal frame interpolation formula is:

[0167]

[0168] Among them, t0 and t1 are two moments with temperature measurements, the subscript "interpolatory" indicates that the temperature is obtained by interpolation, and tk is the moment between t0 and t1 that needs to be interpolated.

[0169] Step 8230, power modeling is performed based on artificial neural network: the data set is split according to weather type, and different power models are trained for different weather types; the input vectors are the actual indoor temperature, the set temperature of the indoor unit, the actual outdoor temperature, and the time; the output variable is the average power after normalization, and the normalization benchmark is the maximum working power of the air conditioner; because the air conditioner power is nonlinear within a limited range, the Sigmoid function is selected as the activation function;

[0170] Among them, the input and output formula of the air conditioning power model is:

[0171]

[0172] The activation function formula of artificial neural network is:

[0173]

[0174] The subscript "rated" indicates the maximum rated value. It is the normalization of power. f refers to the generalized undetermined function. The subscript ANN indicates that the undetermined function is constructed based on a neural network. [] indicates that the input is a set of vectors. The subscript tk indicates that all variables are quantities corresponding to time tk. S(x) is the component function form of the hidden layer in ANN. The right side is the standard expression of Sigmoid, where exp represents the exponential function of natural constants, b is the undetermined constant, and a1 to a n Represents the unknown coefficients, x1 to x n Refers to the variables in [] in the input and output formula of the air conditioning power model.

[0175] Furthermore, when the multi-split variable frequency air conditioner flexible load control program is executed by the processor, the following steps are performed:

[0176] Step S310, judging: based on the air-conditioning power model, under the current indoor parameter state, predict the power range within the comfortable adjustable temperature range. If the power command is within the power range, it is judged that it can be responded. If the power command is outside the power range, it is inquired whether the response residual is allowed. If it is allowed, it responds according to the boundary value of the air-conditioning comfortable temperature range and feeds back the response residual. Otherwise, the response execution is exited. In the cooling mode, the load response judgment formula is:

[0177]

[0178] Among them, the subscript comfort represents the comfort range; up_limit represents the upper limit of the comfort temperature range; dn_limit represents the lower limit of the comfort temperature range; the subscript command represents the command, P command Indicates power command;

[0179] Step S320, selection: based on the air conditioning power model, select the set temperature that is closest to the power command, wherein all settable temperatures are traversed, and the value with the predicted power less than the power command and the smallest load response error is selected, and the partial shutdown strategy is not considered in the initial response. The pseudo code for selecting the set temperature is:

[0180]

[0181] Among them, || indicates absolute value calculation, min indicates taking the smaller value of two calculated values, and Error indicates the response residual; Lists_T set Represents a set of all optional set temperatures for a multi-split variable-frequency air conditioner, configured according to the air conditioner hardware attributes; Ti is an intermediate variable for temperature calculation, T is the selected set temperature value, and return is a command to return to the program;

[0182] Step S330, execution: first send the set temperature to each indoor unit; and continuously monitor the actual power of the air conditioner within the power command control time range, and calculate the sliding average power; the end condition of the execution must meet any one of the following conditions: 1. The power control time ends; 2. The indoor temperature exceeds the comfort temperature range limit for the first preset time period;

[0183] Step S340, tracking feedback: suppress power rebound through tracking feedback, and the criterion for tracking feedback is: starting from the second preset time after the power instruction is issued, if the sliding average power value of the preset number of times in the past third preset time is greater than the power instruction, the suppression strategy is triggered; the set temperature is calculated again according to the air-conditioning power model and the pseudo code for selecting the set temperature; if the set temperature has a value, the set temperature adjustment instruction is issued according to the new set temperature; if the set temperature is empty, an instruction to shut down some indoor units is issued, and the indoor units with low air inlet temperatures are shut down one by one; until the power rebound is suppressed, or the indoor temperature exceeds the comfortable temperature range.

[0184] The beneficial effects of the multi-split frequency conversion air conditioner flexible load control system of the present invention are:

[0185] 1. The algorithm has good adaptability and can adapt to VRV air conditioners of different models and powers and room environments with different thermal capacity and thermal resistance properties; it can automatically adapt to the zero drift of the air conditioner's cooling capacity and changes in the room status, reducing the cost of later debugging and maintenance.

[0186] 2. Minimize the method of directly shutting down the air conditioner, make full use of the flexible load under the set temperature change, and achieve a balance between load response and temperature comfort assurance.

[0187] 3. An innovative method for suppressing air conditioning power rebound is discussed. Based on continuous temperature adjustment and partial shutdown of indoor units, a new relatively balanced cooling load state can be gradually found based on this method, despite the interference of difficulties such as strong thermal inertia and strong heat storage interference in the room.

[0188] To achieve the above-mentioned purpose, the present invention also proposes a computer-readable storage medium, characterized in that the computer-readable storage medium stores a multi-split variable frequency air conditioner flexible load control program, and the multi-split variable frequency air conditioner flexible load control program is executed by the processor when it is run by the processor to execute the steps of the method described in the above embodiment, which will not be repeated here.

[0189] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A flexible load control method for a multi-split variable frequency air conditioner, characterized in that: The method comprises the following steps: Step 1, operation data collection: obtain the real-time total power P of the air conditioner or the outdoor unit power, and the actual indoor temperature Indoor unit set temperature Actual outdoor temperature T out , weather type weather, set temperature change amount and change time tk and corresponding indoor unit number, shutdown number and shutdown time of some indoor units, continuous monitoring data; Among them, the superscript N represents the indoor unit number; the subscript in represents the actual indoor temperature of the room, the subscript set represents the set temperature of the indoor unit of the room, and the subscript out represents the actual outdoor temperature. They are used to distinguish the nature of the temperature; tk is the time mark according to the 24-hour system; Step 2, air conditioning power model construction: build the air conditioning power model based on the collected operation data; Step 3, air conditioning load response: when receiving the power instruction, start the air conditioning load response and execute the flexible load control strategy for the multi-split variable frequency air conditioner, wherein the power instruction includes the allowable power upper limit and power control time, the configuration parameters include the comfort temperature range, the air conditioner settable temperature vector, and the operating parameters include the indoor temperature, outdoor temperature, and time information.

2. The flexible load control method of a multi-split frequency conversion air conditioner according to claim 1, characterized in that: In the step 2, the data set is split according to weather type, and different weather types are used to train different power types. Each time the power model is called, the corresponding weather type must be selected. The step 2 includes: Step S210, power signal filtering processing: filtering the actual power based on the sliding average filtering method, wherein the filtering sliding time is the multi-split variable frequency air conditioner power up and down adjustment cycle, and the sliding average filtering formula of the actual power of the air conditioner is: Among them, Δt represents the power measurement interval, and n represents the power adjustment cycle of the multi-split variable frequency air conditioner; Step S220, temperature signal frame interpolation processing: interpolate the actual temperature frame based on the difference method to ensure that each power has a corresponding actual value of the indoor and outdoor temperature, wherein the temperature signal frame interpolation formula is: Among them, t0 and t1 are two moments with temperature measurements, the subscript interpolatory indicates that the temperature is obtained by interpolation, and tk is the moment between t0 and t1 that needs to be interpolated; Step 8230, power modeling is performed based on artificial neural network: the data set is split according to weather type, and different power models are trained for different weather types; the input vectors are the actual indoor temperature, the set temperature of the indoor unit, the actual outdoor temperature, and the time; the output variable is the average power after normalization, and the normalization benchmark is the maximum working power of the air conditioner; because the air conditioner power is nonlinear within a limited range, the Sigmoid function is selected as the activation function; Among them, the input and output formula of the air conditioning power model is: The activation function formula of artificial neural network is: The subscript "rated" indicates the maximum rated value. It is the normalization of power. f refers to the generalized undetermined function. The subscript ANN indicates that the undetermined function is constructed based on a neural network. [] indicates that the input is a set of vectors. The subscript tk indicates that all variables are quantities corresponding to time tk. S(x) is the component function form of the hidden layer in ANN. The right side is the standard expression of Sigmoid, where exp represents the exponential function of natural constants, b is the undetermined constant, and a1 to a n Represents the unknown coefficients, x1 to x n Refers to the variables in [] in the input and output formula of the air conditioning power model.

3. The flexible load control method of a multi-split frequency conversion air conditioner according to claim 2, characterized in that: The step 3 includes: Step S310, judging: based on the air-conditioning power model, under the current indoor parameter state, predict the power range within the comfortable adjustable temperature range. If the power command is within the power range, it is judged that it can be responded. If the power command is outside the power range, it is inquired whether the response residual is allowed. If it is allowed, it responds according to the boundary value of the air-conditioning comfortable temperature range and feeds back the response residual. Otherwise, the response execution is exited. In the cooling mode, the load response judgment formula is: Among them, the subscript comfort represents the comfort range; up_limit represents the upper limit of the comfort temperature range; dn_limit represents the lower limit of the comfort temperature range; the subscript command represents the command, P command Indicates power command; Step 8320, selection: based on the air conditioning power model, select the set temperature that is closest to the power command, wherein all settable temperatures are traversed, and the value with the predicted power less than the power command and the smallest load response error is selected, and the partial shutdown strategy is not considered in the initial response. The pseudo code for selecting the set temperature is: Among them, || indicates absolute value calculation, min indicates taking the smaller value of two calculated values, and Error indicates the response residual; Lists_T set Represents a set of all optional set temperatures for a multi-split variable-frequency air conditioner, configured according to the air conditioner hardware attributes; Ti is an intermediate variable for temperature calculation, T is the selected set temperature value, and return is a command to return to the program; Step S330, execution: first send the set temperature to each indoor unit; and continuously monitor the actual power of the air conditioner within the power command control time range, and calculate the sliding average power; the end condition of the execution must meet any one of the following conditions:

1. The power control time ends; 2. The indoor temperature exceeds the comfort temperature range limit for the first preset time period; Step S340, tracking feedback: suppress power rebound through tracking feedback, and the criterion for tracking feedback is: starting from the second preset time after the power instruction is issued, if the sliding average power value of the preset number of times in the past third preset time is greater than the power instruction, the suppression strategy is triggered; the set temperature is calculated again according to the air-conditioning power model and the pseudo code for selecting the set temperature; if the set temperature has a value, the set temperature adjustment instruction is issued according to the new set temperature; if the set temperature is empty, an instruction to shut down some indoor units is issued, and the indoor units with low air inlet temperatures are shut down one by one; until the power rebound is suppressed, or the indoor temperature exceeds the comfortable temperature range.

4. A flexible load control system for a multi-split frequency conversion air conditioner, characterized in that: The system includes a memory, a processor, and a multi-split variable frequency air conditioner flexible load control program stored on the processor. When the multi-split variable frequency air conditioner flexible load control program is run by the processor, the following steps are performed: Step 1, operation data collection: obtain the real-time total power P of the air conditioner or the outdoor unit power, and the actual indoor temperature Indoor unit set temperature Actual outdoor temperature T out , weather type weather, set temperature change amount and change time tk and corresponding indoor unit number, shutdown number and shutdown time of some indoor units, continuous monitoring data; Among them, the superscript N represents the indoor unit number; the subscript in represents the actual indoor temperature of the room, the subscript set represents the set temperature of the indoor unit of the room, and the subscript out represents the actual outdoor temperature. They are used to distinguish the nature of the temperature; tk is the time mark according to the 24-hour system; Step 2, air conditioning power model construction: build the air conditioning power model based on the collected operation data; Step 3, air conditioning load response: when receiving the power instruction, start the air conditioning load response and execute the flexible load control strategy for the multi-split variable frequency air conditioner, wherein the power instruction includes the allowable power upper limit and power control time, the configuration parameters include the comfort temperature range, the air conditioner settable temperature vector, and the operating parameters include the indoor temperature, outdoor temperature, and time information.

5. The multi-split variable frequency air conditioner flexible load control system according to claim 4 is characterized in that: When the multi-split variable frequency air conditioner flexible load control program is executed by the processor, the following steps are further performed: Step S210, power signal filtering processing: filtering the actual power based on the sliding average filtering method, wherein the filtering sliding time is the multi-split variable frequency air conditioner power up and down adjustment cycle, and the sliding average filtering formula of the actual power of the air conditioner is: Among them, Δt represents the power measurement interval, and n represents the power adjustment cycle of the multi-split variable frequency air conditioner; Step S220, temperature signal frame interpolation processing: interpolate the actual temperature frame based on the difference method to ensure that each power has a corresponding actual value of the indoor and outdoor temperature, wherein the temperature signal frame interpolation formula is: Among them, t0 and t1 are two moments with temperature measurements, the subscript interpolatory indicates that the temperature is obtained by interpolation, and tk is the moment between t0 and t1 that needs to be interpolated; Step S230, power modeling is performed based on artificial neural network: the data set is split according to weather type, and different power models are trained for different weather types; the input vectors are the actual indoor temperature, the set temperature of the indoor unit, the actual outdoor temperature, and the time; the output variable is the average power after normalization, and the normalization benchmark is the maximum working power of the air conditioner; since the air conditioner power is nonlinear within a limited range, the Sigmoid function is selected as the activation function; Among them, the input and output formula of the air conditioning power model is: The activation function formula of artificial neural network is: The subscript "rated" indicates the maximum rated value. It is a standardized processing of power. f refers to a generalized undetermined function. The subscript ANN indicates that this undetermined function is constructed based on a neural network. [] indicates that the input quantity is a set of vectors. The subscript tk indicates that all variables are quantities corresponding to time tk. S(x) is the component function form of the hidden layer in ANN. is the standard expression of Sigmoid, where exp represents the exponential function of a natural constant, b is an undetermined constant, and a1 to a n Represents the unknown coefficients, x1 to x n Refers to the variables in [] in the input and output formula of the air conditioning power model.

6. The flexible load control method of a multi-split variable frequency air conditioner according to claim 5, characterized in that: The multi-split variable frequency air conditioner flexible load control program is executed by the processor to perform the following steps: Step S310, judging: based on the air-conditioning power model, under the current indoor parameter state, predict the power range within the comfortable adjustable temperature range. If the power command is within the power range, it is judged that it can be responded. If the power command is outside the power range, it is inquired whether the response residual is allowed. If it is allowed, it responds according to the boundary value of the air-conditioning comfortable temperature range and feeds back the response residual. Otherwise, the response execution is exited. In the cooling mode, the load response judgment formula is: Among them, the subscript comfort represents the comfort range; up-limit represents the upper limit of the comfort temperature range; dn-limit represents the lower limit of the comfort temperature range; the subscript command represents the command, P command Indicates power command; Step S320, selection: based on the air conditioning power model, select the set temperature that is closest to the power command, wherein all settable temperatures are traversed, and the value with the predicted power less than the power command and the smallest load response error is selected, and the partial shutdown strategy is not considered in the initial response. The pseudo code for selecting the set temperature is: Among them, || indicates absolute value calculation, min indicates taking the smaller value of two calculated values, and Error indicates the response residual; Lists-T set Represents a set of all optional set temperatures for a multi-split variable-frequency air conditioner, configured according to the air conditioner hardware attributes; Ti is an intermediate variable for temperature calculation, T is the selected set temperature value, and return is a command to return to the program; Step S330, execution: first send the set temperature to each indoor unit; and continuously monitor the actual power of the air conditioner within the power command control time range, and calculate the sliding average power; the end condition of the execution must meet any one of the following conditions:

1. The power control time ends; 2. The indoor temperature exceeds the comfort temperature range limit for the first preset time period; Step S340, tracking feedback: suppress power rebound through tracking feedback, and the criterion for tracking feedback is: starting from the second preset time after the power instruction is issued, if the sliding average power value of the preset number of times in the past third preset time is greater than the power instruction, the suppression strategy is triggered; the set temperature is calculated again according to the air-conditioning power model and the pseudo code for selecting the set temperature; if the set temperature has a value, the set temperature adjustment instruction is issued according to the new set temperature; if the set temperature is empty, an instruction to shut down some indoor units is issued, and the indoor units with low air inlet temperatures are shut down one by one; until the power rebound is suppressed, or the indoor temperature exceeds the comfortable temperature range.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a multi-split variable frequency air conditioner flexible load control program, and the multi-split variable frequency air conditioner flexible load control program is executed by a processor to perform the steps of the method described in any one of claims 1 to 3.

Images