Air conditioning equipment, control method thereof and computer readable storage medium

Abstract

The invention discloses air conditioning equipment, a control method of the air conditioning equipment and a computer readable storage medium, and relates to the technical field of the air conditioning equipment, and the control method of the air conditioning equipment comprises the steps that multiple candidate parameter values of control parameters of the air conditioning equipment in the next operation period are obtained, and the multiple candidate parameter values of the control parameters of the air conditioning equipment in the next operation period are obtained based on a feedforward control strategy and a feedback control strategy; predicting a control deviation value of each control target of the air conditioning equipment under each candidate parameter value of the control parameters; based on the target weight value of each control target, performing weighted fusion processing on the control deviation value of each control target under each candidate parameter value to obtain a comprehensive deviation value of each candidate parameter value; and the candidate parameter value with the minimum comprehensive deviation value in all the candidate parameter values serves as a target parameter value, and the air conditioning equipment is controlled to operate according to the target parameter value of the control parameter in the next operation period. The control effect and the control adaptability of the air conditioning equipment can be improved at the same time.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of air conditioning equipment, in particular to an air conditioning equipment, a control method thereof and a computer readable storage medium. BACKGROUND

[0002] At present, PID (Proportion Integral Differential) frequency conversion technology is usually adopted to realize the control of air conditioning equipment. However, the control target of this control technology is single, and it is difficult to realize multi-target collaborative control, thereby affecting the control effect; and this control technology is prone to produce a large response delay in the control process, and is difficult to adapt to the dynamic variable real environment, thereby causing poor control adaptability. SUMMARY

[0003] The main purpose of the present application is to provide an air conditioning equipment, a control method thereof and a computer readable storage medium, aiming to improve the control effect and control adaptability of the air conditioning equipment at the same time.

[0004] To achieve the above purpose, the present application provides a control method of an air conditioning equipment, which comprises: obtaining a plurality of candidate parameter values of a control parameter of an air conditioning equipment in a next operation cycle, and predicting, based on a feedforward control strategy and a feedback control strategy, a control deviation value of each control target of the air conditioning equipment under each candidate parameter value of the control parameter; the control target is used to represent a performance index that the air conditioning equipment needs to achieve, and the control deviation value is used to represent the degree to which the control target meets a set control target requirement value; obtaining a target weight value of each control target under the current control requirement of the air conditioning equipment; based on the target weight value of each control target, performing weighted fusion processing on the control deviation value of each control target under each candidate parameter value to obtain a comprehensive deviation value of each candidate parameter value; taking the candidate parameter value with the minimum comprehensive deviation value in each candidate parameter value as a target parameter value, and controlling the air conditioning equipment to operate according to the target parameter value of the control parameter in the next operation cycle.

[0005] In an embodiment, the step of predicting, based on a feedforward control strategy and a feedback control strategy, a control deviation value of each control target of the air conditioning equipment under each candidate parameter value of the control parameter comprises: obtaining a comprehensive prediction model coupled with the feedforward control strategy and the feedback control strategy, and obtaining current state data of the air conditioning equipment; inputting the current state data and each of the candidate parameter values of the control parameter into the comprehensive prediction model to obtain a first prediction value of each of the control targets under each of the candidate parameter values of the control parameter; calculating a difference between a demand value of each of the control targets and the first prediction value of each of the control targets under each of the candidate parameter values to obtain a control deviation value of each of the control targets under each of the candidate parameter values.

[0006] In an embodiment, the step of obtaining the comprehensive prediction model coupled with the feedforward control strategy and the feedback control strategy comprises: obtaining a difference between a demand value and an actual value of each of the control targets in a plurality of historical operation periods before a current operation period to obtain a correction loss of each of the control targets in the plurality of historical operation periods; constructing a training sample set according to the correction loss of each of the control targets in the plurality of historical operation periods; retraining a pre-trained feedforward prediction model according to the training sample set to generate the comprehensive prediction model.

[0007] In an embodiment, the step of retraining the pre-trained feedforward prediction model according to the training sample set to generate the comprehensive prediction model comprises: freezing a preset number of continuous network layers in the feedforward prediction model; performing transfer learning on the network layers that are not frozen in the feedforward prediction model according to the training sample set to generate the comprehensive prediction model.

[0008] In an embodiment, the step of retraining the pre-trained feedforward prediction model according to the training sample set to generate the comprehensive prediction model further comprises: performing fine-tuning training on part of the network parameters in the feedforward prediction model according to the training sample set to generate the comprehensive prediction model.

[0009] In an embodiment, for an environmental target in each of the control targets, the comprehensive prediction model comprises a multi-task expert network model; The step of inputting the current state data and each of the candidate parameter values of the control parameter into the comprehensive prediction model to obtain a first prediction value of each of the control targets under each of the candidate parameter values of the control parameter comprises: For any of the candidate parameter values of the control parameter, the current state data and the candidate parameter value are input into the multi-task expert network model; Based on each expert network in the multi-task expert network model, the current state data and the candidate parameter value are subjected to feature extraction processing to obtain data processed by each expert network; each expert network includes a temperature expert network, a humidity expert network, an air volume expert network, and a public network; Based on the gating structure in the multi-task expert network model, a self-balancing training mechanism is adopted to fuse the data processed by each expert network to obtain fused data; The fused data is input into each discriminant network in the multi-task expert network model to obtain a temperature prediction value, a humidity prediction value, and an air volume prediction value, which are collectively used as the first prediction value of the environmental target under the candidate parameter value.

[0010] In an embodiment, for each energy-saving target in the control target, the comprehensive prediction model includes a sensible heat ratio prediction model and an energy efficiency prediction model; The step of inputting the current state data and each candidate parameter value of the control parameter into the comprehensive prediction model to obtain the first prediction value of each control target under each candidate parameter value of the control parameter includes: For any candidate parameter value of the control parameter, the current state data and the candidate parameter value are input into the input layer of the sensible heat ratio prediction model; Based on the processing of the feature extraction layer and the temperature and humidity energy distribution layer of the sensible heat ratio prediction model, a first sensible heat ratio of a next operation cycle and a second sensible heat ratio of a current operation cycle are obtained; In the output layer of the sensible heat ratio prediction model, the first sensible heat ratio and the second sensible heat ratio are subjected to constraint processing to obtain a predicted sensible heat ratio of the next operation cycle; And the current state data and the candidate parameter value are input into the energy efficiency prediction model; Based on the processing of the low-dimensional feature extraction layer and the high-dimensional feature mapping layer of the energy efficiency prediction model, a high-dimensional feature matrix is obtained; In the energy efficiency prediction layer of the energy efficiency prediction model, the high-dimensional feature matrix is subjected to constraint processing to obtain a predicted energy efficiency; The predicted sensible heat ratio and the predicted energy efficiency are collectively used as the first prediction value of the energy-saving target under the candidate parameter value.

[0011] In an embodiment, for each comfort target in the control target, the comprehensive prediction model includes a comfort parameter prediction model; The step of inputting the current state data and each candidate parameter value of the control parameter into the comprehensive prediction model to obtain the first prediction value of each control target under each candidate parameter value of the control parameter includes: inputting the current state data and the candidate parameter value of the control parameter into the comfort parameter prediction model to obtain a human parameter prediction value and a thermal comfort prediction value, wherein the comfort parameter prediction model comprises a first model trained with a distribution dispersion degree of human parameters before and after switching of the running state of the air conditioning equipment as a constraint condition, and a second model trained with a difference in absolute value of the thermal comfort value before and after switching of the running state of the air conditioning equipment as a constraint condition; taking the human parameter prediction value and the thermal comfort prediction value as a first prediction value of the comfort target under the candidate parameter value.

[0012] In an embodiment, the step of predicting, based on the feedforward control strategy and the feedback control strategy, a control deviation value of each control target of the air conditioning equipment under each candidate parameter value of the control parameter further comprises: obtaining current state data of the air conditioning equipment; inputting the current state data and each candidate parameter value of the control parameter into a feedforward prediction model associated with the feedforward control strategy to obtain a second prediction value of each control target under each candidate parameter value; calculating a difference between a demand value of each control target set and the second prediction value of each control target under each candidate parameter value to obtain a first deviation value of each control target under each candidate parameter value; obtaining a third prediction value of each control target under a current parameter value of the control parameter predicted by a feedback prediction model associated with the feedback control strategy in a previous running period; calculating a difference between an actual value of each control target in the current running period and the third prediction value of each control target to obtain a second deviation value of each control target; performing weighted fusion processing on the first deviation value and the second deviation value of each control target under each candidate parameter value to obtain a control deviation value of each control target under each candidate parameter value.

[0013] In an embodiment, the step of constructing the feedback prediction model comprises: obtaining a difference between a demand value and an actual value of each control target in a plurality of historical running periods before the current running period to obtain a correction loss of each control target in the plurality of historical running periods; constructing a training sample set according to the correction loss of each control target in the plurality of historical running periods; performing reinforcement learning on a to-be-trained model according to the training sample set to generate the feedback prediction model.

[0014] In an embodiment, the step of obtaining the current state data of the air conditioning device comprises: obtaining a compressor current energy of a current operation cycle of the air conditioning device and compressor historical energies of a plurality of historical operation cycles before the current operation cycle; obtaining an inertial energy of each historical operation cycle transferred to the current operation cycle according to the compressor historical energies of the plurality of historical operation cycles; obtaining a cumulative energy of the current operation cycle according to the inertial energies and the compressor current energy; determining the current state data of the air conditioning device based on the cumulative energy.

[0015] In an embodiment, the step of obtaining the inertial energy of each historical operation cycle transferred to the current operation cycle according to the compressor historical energies of the plurality of historical operation cycles comprises: obtaining an energy transfer inertia coefficient of the compressor historical energy of each historical operation cycle transferred to the current operation cycle; calculating the inertial energy of each historical operation cycle transferred to the current operation cycle according to the compressor historical energy of each historical operation cycle and the energy transfer inertia coefficient of the compressor historical energy of each historical operation cycle transferred to the current operation cycle.

[0016] In an embodiment, the step of determining the current state data of the air conditioning device based on the cumulative energy comprises: obtaining state data corresponding to the cumulative energy as the current state data based on a preset mapping relationship between compressor energy and state data.

[0017] In an embodiment, the control targets comprise an environment target, a comfort target and an energy saving target, and the step of obtaining target weight values of the control targets under current control requirements of the air conditioning device comprises: calculating a time interval between the current operation cycle and a time when a user last set requirement values of the control targets and obtaining a current electricity price; determining an initial weight value of the environment target and an initial weight value of the comfort target according to the time interval; the initial weight value of the environment target is negatively related to the time interval, and the initial weight value of the comfort target is positively related to the time interval; determining an initial weight value of the energy saving target according to the electricity price; the initial weight value of the energy saving target is positively related to the electricity price; The initial weight value of the environment target, the initial weight value of the comfort target and the initial weight value of the energy saving target are normalized to obtain a target weight value of the environment target, a target weight value of the comfort target and a target weight value of the energy saving target.

[0018] In an embodiment, the step of obtaining the target weight value of each control target under the current control requirement of the air conditioning device further comprises: Based on a preset mapping relationship between the running function and the weight value of each control target, the weight value of each control target corresponding to the current running function of the air conditioning device is obtained as the target weight value of each control target.

[0019] In an embodiment, the step of obtaining the plurality of candidate parameter values of the control parameter of the air conditioning device comprises: obtaining a historical parameter value of the control parameter in the last running cycle; determining the plurality of candidate parameter values of the control parameter according to the historical parameter value.

[0020] In addition, to achieve the above object, the present application also provides an air conditioning device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the computer program is configured to implement the steps of the control method of the air conditioning device.

[0021] In addition, to achieve the above object, the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the control method of the air conditioning device.

[0022] In addition, to achieve the above object, the present application also provides a computer program product, which comprises a computer program, and the computer program is executed by a processor to implement the steps of the control method of the air conditioning device.

[0023] The present application provides a control method for air-conditioning equipment, which obtains multiple candidate parameter values ​​of the control parameters of the air-conditioning equipment in the next operating cycle, and predicts the control deviation value of each control target of the air-conditioning equipment under each candidate parameter value of the control parameter based on a feedforward control strategy and a feedback control strategy; the control target is used to characterize the performance index to be achieved by the air-conditioning equipment, and the control deviation value is used to characterize the extent to which the control target meets the set control target requirement value; the target weight value of each control target under the current control requirement of the air-conditioning equipment is obtained; based on the target weight value of each control target, the control deviation value of each control target under each candidate parameter value is weightedly fused to obtain a comprehensive deviation value of each candidate parameter value; the candidate parameter value with the smallest comprehensive deviation value among the candidate parameter values ​​is used as the target parameter value, and the air-conditioning equipment is controlled to operate according to the target parameter value of the control parameter in the next operating cycle.

[0024] Therefore, the technical solution provided by this application determines the target parameter values ​​of the control parameters of the air conditioning equipment in the next operating cycle by simultaneously combining the feedforward control strategy, the feedback control strategy, and the target weight value of each control target under the current control demand of the air conditioning equipment, so that each control target satisfies its own control target demand value as much as possible, thereby achieving multi-objective coordinated control of the air conditioning equipment with better control effects. Furthermore, because the technical solution provided by this application adopts a feedforward control strategy with a priori effect in the process of achieving multi-objective coordinated control, the technical solution provided by this application can adapt to dynamic and variable real environments.

[0025] In summary, the technical solution provided in this application can simultaneously improve the control effect and control adaptability of air conditioning equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0027] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0028] Figure 1 A schematic diagram illustrating the overall implementation of the control method for air conditioning equipment provided in an embodiment of the present application; Figure 2 A schematic flow chart of a control method for an air conditioning device according to a first embodiment of the present application; Figure 3A working principle diagram of the multi-task expert network model provided by the third embodiment of the present application is shown in the figure; Figure 4 An implementation principle diagram of the sensible heat ratio prediction model provided by the fourth embodiment of the present application is shown in the figure; Figure 5 An architecture diagram of the comfort parameter prediction model provided by the fifth embodiment of the present application is shown in the figure; Figure 6 An implementation principle diagram of the compressor energy accumulated in multiple running cycles provided by the seventh embodiment of the present application is shown in the figure; Figure 7 A structure schematic diagram of the hardware running environment involved in the embodiments of the present application is shown in the figure.

[0029] The object implementation, functional features and advantages of the present application will be further explained with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0030] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application, and are not used to limit the present application.

[0031] In order to better understand the technical solutions of the present application, the following will be described in detail in combination with the drawings of the specification and specific embodiments.

[0032] At present, PID variable frequency technology is usually used to realize the control of air conditioning equipment. However, the control target of this control technology is single, and it is difficult to realize multi-target collaborative control, thereby affecting the control effect; and this control technology is prone to produce large response delay in the process of control, and is difficult to adapt to the dynamic variable real environment, thereby causing poor control adaptability.

[0033] In addition, this control technology is subject to its control granularity and control mechanism, and it is difficult to handle nonlinear complex systems, and is prone to produce large control error, thereby affecting the control precision, resulting in over-temperature, under-temperature, temperature oscillation and other phenomena of air conditioning equipment. And this control technology is difficult to achieve good energy saving effect and comfort effect.

[0034] Based on this, the application provides a control method of an air conditioning device, obtaining a plurality of candidate parameter values of a control parameter of the air conditioning device in a next operation cycle, and based on a feedforward control strategy and a feedback control strategy, predicting a control deviation value of each control target of the air conditioning device under each candidate parameter value of the control parameter; the control target is used to represent a performance index that the air conditioning device needs to achieve, and the control deviation value is used to represent a degree to which the control target meets a set control target requirement value; obtaining a target weight value of each control target under a current control requirement of the air conditioning device; based on the target weight value of each control target, performing a weighted fusion processing on the control deviation value of each control target under each candidate parameter value, to obtain a comprehensive deviation value of each candidate parameter value; taking the candidate parameter value with the minimum comprehensive deviation value in each candidate parameter value as a target parameter value, and controlling the air conditioning device to operate according to the target parameter value of the control parameter in the next operation cycle.

[0035] Therefore, the technical scheme provided by the application can determine the target parameter value of the control parameter of the air conditioning device in the next operation cycle by combining the feedforward control strategy, the feedback control strategy and the target weight value of each control target under the current control requirement of the air conditioning device, so that each control target can meet the respective control target requirement value as much as possible, thereby realizing the multi-target collaborative control of the air conditioning device with better control effect. Moreover, since the technical scheme provided by the application adopts the feedforward control strategy with a priori effect in the process of realizing the multi-target collaborative control, the technical scheme provided by the application can adapt to the dynamic and variable real environment.

[0036] In summary, the technical scheme provided by the application can improve the control effect and the control adaptability of the air conditioning device at the same time.

[0037] In addition, in the process of predicting the prediction value of each control target, the technical scheme provided by the application not only considers the current energy of the compressor in the current operation cycle, but also considers the historical energy of the compressor in a plurality of historical operation cycles before the current operation cycle, which is respectively transmitted to the inertial energy of the current operation cycle, so as to effectively solve the influence of the control accuracy of the air conditioning device caused by the existence of the energy transmission inertia, thereby improving the control accuracy of the air conditioning device.

[0038] The overall implementation principle of the control method of the air conditioning device provided by the application mentioned above can refer to the description of the control method of the air conditioning device provided by the application. Figure 1Specifically, the application determines the prediction loss of each control target by using the prediction value and the demand value of each control target (i.e., the multi-target prediction value and the multi-target demand value), thereby constructing a feedforward control strategy; and determines the correction loss of each control target by using the demand value and the actual value of each control target (i.e., the multi-target demand value and the multi-target actual value), thereby constructing a feedback control strategy. Thus, the application will simultaneously use the feedforward control strategy and the feedback control strategy to analyze the prediction value of each control target under the control cycle of the air conditioning device, so as to reasonably determine the target parameter value of the control parameter, thereby improving the control effect and the control adaptability of the air conditioning device. Moreover, the application also considers the influence of the energy transmission inertia on the prediction value of each control target, so as to solve the control hysteresis problem, thereby improving the control accuracy of the air conditioning device.

[0039] The actual value of the control target is the value actually achieved by the control target, the prediction value of the control target is the value that the control target will theoretically achieve according to the prediction, and the demand value of the control target is the value that the control target needs to achieve. The prediction loss is the difference between the demand value and the prediction value of the control target, and the correction loss is the difference between the demand value and the actual value of the control target.

[0040] The execution subject of the control method of the air conditioning device can be the air conditioning device, or a device that adjusts air by controlling other devices, such as a server, a central controller, a central control device, a line control device, and the like, or a control system or a control circuit with corresponding functions, which are not specifically limited in the embodiment.

[0041] The following embodiments are described by taking the air conditioning device as an execution subject.

[0042] Based on this, the application proposes a control method of an air conditioning device according to the first embodiment, which is described as follows. Figure 2 The control method of the air conditioning device includes steps S10-S40. In step S10, the multiple candidate parameter values of the control parameter of the air conditioning device in the next operation cycle are obtained, and the control deviation values of each control target of the air conditioning device under each candidate parameter value of the control parameter are predicted based on the feedforward control strategy and the feedback control strategy; the control target is used to represent the performance index that the air conditioning device needs to achieve, and the control deviation value is used to represent the degree to which the control target meets the demand value of the control target. It should be noted that the control parameter can include, but is not limited to, fan speed, compressor frequency, and / or expansion valve opening degree, and the like, and the embodiment is not limited specifically in this regard. The candidate parameter value refers to a parameter value of the air conditioning device that can be selected to achieve in the next operation cycle. The multiple control targets of the air conditioning device can include environmental targets, energy saving targets, and comfort targets, and can also include multiple indexes under the environmental targets (such as temperature, humidity, and air volume), can also include multiple indexes under the energy saving targets (such as capacity, energy efficiency, and energy consumption), and can also include multiple indexes under the comfort targets (such as PMV (Predicted Mean Vote) and UIC (User's Individual Comfort)), and the embodiment is not limited specifically in this regard.

[0043] In addition, it should be noted that when the multiple candidate parameter values of the control parameter of the air conditioning device in the next operation cycle are obtained, the candidate parameter values can be determined by the user according to experience, all the parameter values set can be directly used as the candidate parameter values, or the candidate parameter values can be determined according to the historical parameter values of the control parameter in the last operation cycle, and the embodiment is not limited specifically in this regard.

[0044] When the multiple candidate parameter values are determined according to the historical parameter values of the control parameter in the last operation cycle, the parameter value range can be determined based on the historical parameter values, the preset maximum parameter change value, and the preset minimum parameter change value, and all or part of the parameter values in the parameter value range can be used as the candidate parameter values.

[0045] In step S20, target weight values of the control targets under the current control requirement of the air conditioning device are obtained. It should be noted that the weight value of the control target under the current control requirement of the air conditioning device is referred to as a target weight value in the embodiment.

[0046] In a feasible implementation, when the control targets include environmental targets, comfort targets, and energy saving targets, step S20 can include steps S21-S24. In step S21, a time interval between the current operation cycle and a time when the user last set the requirement value of each control target is calculated, and a current electricity price is obtained. In step S22, initial weight values of the environmental targets and initial weight values of the comfort targets are determined according to the time interval; the initial weight values of the environmental targets are negatively related to the time interval, and the initial weight values of the comfort targets are positively related to the time interval. It should be noted that the initial weight value of the environmental target is negatively related to the time interval, that is, the greater the time interval, the smaller the initial weight value of the environmental target; the smaller the time interval, the greater the initial weight value of the environmental target. The initial weight value of the comfort target is positively related to the time interval, that is, the greater the time interval, the greater the initial weight value of the comfort target; the smaller the time interval, the smaller the initial weight value of the comfort target.

[0047] In determining the initial weight value of the environmental target and the initial weight value of the comfort target according to the time interval, the weight value of the environmental target corresponding to the time interval can be obtained as the initial weight value of the environmental target based on the mapping relationship between the preset time interval and the weight value of the environmental target, and the weight value of the comfort target corresponding to the time interval can be obtained as the initial weight value of the comfort target based on the mapping relationship between the preset time interval and the weight value of the comfort target.

[0048] Among them, the mapping relationship between the time interval and the weight value of the environmental target, and the mapping relationship between the time interval and the weight value of the comfort target can be recorded in the form of a relationship table, a relationship curve or a key-value pair, and the present embodiment does not make specific limitation thereto.

[0049] Step S23, determining the initial weight value of the energy saving target according to the electricity price; the initial weight value of the energy saving target is positively related to the electricity price; It should be noted that the initial weight value of the energy saving target is positively related to the electricity price, that is, the higher the electricity price, the greater the weight value of the energy saving target; the lower the electricity price, the smaller the weight value of the energy saving target.

[0050] In determining the initial weight value of the energy saving target according to the electricity price, the weight value of the energy saving target corresponding to the electricity price can be obtained as the initial weight value of the energy saving target based on the mapping relationship between the preset electricity price and the weight value of the energy saving target.

[0051] Among them, the mapping relationship between the electricity price and the weight value of the energy saving target can be recorded in the form of a relationship table, a relationship curve or a key-value pair, and the present embodiment does not make specific limitation thereto.

[0052] Step S24, normalizing the initial weight value of the environmental target, the initial weight value of the comfort target and the initial weight value of the energy saving target to obtain the target weight value of the environmental target, the target weight value of the comfort target and the target weight value of the energy saving target.

[0053] It should be noted that when the initial weight value of the environment target, the initial weight value of the comfort target and the initial weight value of the energy saving target are normalized to obtain the target weight value of the environment target, the target weight value of the comfort target and the target weight value of the energy saving target, the sum of the initial weight value of the environment target, the initial weight value of the comfort target and the initial weight value of the energy saving target can be calculated first to obtain the total weight value; and then the ratio of each of the initial weight value of the environment target, the initial weight value of the comfort target and the initial weight value of the energy saving target to the total weight value is calculated to obtain the target weight value of the environment target, the target weight value of the comfort target and the target weight value of the energy saving target.

[0054] By normalizing the initial weight value of the environment target, the initial weight value of the comfort target and the initial weight value of the energy saving target, the initial weight values of the three targets can be normalized to the same data dimension, so that the target weight value of the environment target, the target weight value of the comfort target and the target weight value of the energy saving target are obtained, and the sum of the three weight values is 1.

[0055] It can be understood that in the actual operation of the air conditioning device, in general, the longer the air conditioning device operates, the higher the user's attention to the comfort target and the lower the user's attention to the environment target; and the higher the electricity price, the higher the user's attention to the energy saving target. Based on this, the present embodiment can accurately and reasonably determine the target weight value of each of the environment target, the comfort target and the energy saving target under the current control demand of the air conditioning device by using the time interval between the current operation period of the air conditioning device and the time when the user last set the demand value of each control target and the current electricity price.

[0056] In another possible implementation, step S20 can include step S25: In step S25, based on the preset mapping relationship between the operation function and the weight value of each control target, the weight value of each control target corresponding to the current operation function of the air conditioning device is obtained as the target weight value of each control target.

[0057] It should be noted that the mapping relationship between the operation function and the weight value of each control target can be recorded in the form of a relationship table, a key-value pair, etc., and the present embodiment does not make specific limitation thereon. The current operation function is the function of the air conditioning device at the current time, and the operation function can be a refrigeration function, a heating function, a dehumidification function, etc., and the present embodiment does not make specific limitation thereon.

[0058] The present embodiment accurately and reasonably determines the target weight value of each control target under the current control demand of the air conditioning device by using the current operation function of the air conditioning device to fit the actual operation condition of the air conditioning device.

[0059] The above are only two possible implementations of step S20 provided by the embodiment, and the embodiment does not specifically limit the specific implementation of step S20.

[0060] In step S30, based on the target weight values of the control targets, the control deviation values of the control targets under each candidate parameter value are weighted and fused to obtain a comprehensive deviation value of each candidate parameter value. It should be noted that when the control deviation values of the control targets under each candidate parameter value are weighted and fused based on the target weight values of the control targets to obtain a comprehensive deviation value of each candidate parameter value, for any candidate parameter value, the product of the control deviation values of the control targets under the candidate parameter value and the target weight values of the control targets is calculated to obtain the weighted deviation values of the control targets. Then, the sum of the weighted deviation values of the control targets is calculated, i.e., the comprehensive deviation value of the candidate parameter value is obtained. The specific implementation process can be referred to formula 1 as follows.

[0061] Formula 1 Wherein, is a comprehensive deviation value of a candidate parameter value, ~ is a target weight value of a control target, ~ is a control deviation value of a control target under a candidate parameter value.

[0062] In step S40, the candidate parameter value with the minimum comprehensive deviation value among the candidate parameter values is taken as the target parameter value, and the air conditioning equipment is controlled to operate according to the target parameter value of the control parameter in the next operation period.

[0063] It should be noted that the smaller the comprehensive deviation value of the candidate parameter value, the higher the degree to which the control parameter of the air conditioning equipment is operated according to the candidate parameter value, and the higher the degree to which the control targets of the air conditioning equipment meet the corresponding demand values.

[0064] As can be seen from the above, the technical solution provided by the embodiment determines the target parameter value of the control parameter of the air conditioning equipment in the next operation period by simultaneously combining the feedforward control strategy, the feedback control strategy, and the target weight value of each control target under the current control demand of the air conditioning equipment, so that each control target meets the respective control target demand value as much as possible, thereby achieving multi-target collaborative control of the air conditioning equipment with better control effect. Moreover, since the technical solution provided by the embodiment adopts the feedforward control strategy with a priori effect in the process of implementing multi-target collaborative control, the technical solution provided by the embodiment can adapt to a dynamically variable real environment.

[0065] In conclusion, the technical solution provided by the embodiment can improve the control effect and control adaptability of the air conditioning equipment.

[0066] Based on the first embodiment, a second embodiment of the control method of the air conditioning equipment is provided, in which step S10 can include steps S11-S13. In step S11, a comprehensive prediction model coupled with the feedforward control strategy and the feedback control strategy is obtained, and current state data of the air conditioning equipment is obtained. It should be noted that the current state data is the state data of the air conditioning equipment in the current operation period, which can include but is not limited to temperature, humidity, fan speed and / or compressor frequency, and other data reflecting the operating state of the air conditioning equipment, which is not limited in the embodiment.

[0067] In a possible implementation, step S11 can include steps S111-S113. In step S111, the difference between the demand value and the actual value of each control target in a plurality of historical operation periods before the current operation period is obtained, and the correction loss of each control target in the plurality of historical operation periods is obtained. In step S112, a training sample set is constructed according to the correction loss of each control target in the plurality of historical operation periods. It should be noted that when the training sample set is constructed according to the correction loss of each control target in the plurality of historical operation periods, the state data and the parameter values of the control parameters of the air conditioning equipment in the historical operation period corresponding to the correction loss can be used as input features, and the correction loss can be used as a sample label, so as to construct the training sample set.

[0068] In step S113, the pre-trained feedforward prediction model is retrained according to the training sample set, and a comprehensive prediction model is generated.

[0069] It should be noted that when the feedforward prediction model is retrained according to the training sample set, the training sample set can be divided into a training set and a validation set according to a certain proportion, so that the training set is used to retrain the feedforward prediction model, and the validation set is used to verify whether the retrained model converges, and after the retrained model converges, the comprehensive prediction model can be generated.

[0070] In this embodiment, the difference between the demand value and the actual value of each control target in a plurality of historical running periods before the current running period is obtained to determine a plurality of losses of each control target under the feedback mechanism, i.e., correction losses; then the correction losses of each control target in the plurality of historical running periods are used to construct a training sample set; and then the pre-trained feedforward prediction model is retrained by using the training sample set, so that the feedback error information (i.e., the correction losses) is integrated into the feedforward prediction model, and finally a comprehensive prediction model coupled with the feedforward control strategy and the feedback control strategy is generated.

[0071] Further, in a possible implementation, the step S113 can include steps S101-S102: In step S101, a preset number of continuous network layers in the feedforward prediction model are frozen. In step S102, the network layers that are not frozen in the feedforward prediction model are subjected to transfer learning according to the training sample set, to generate a comprehensive prediction model.

[0072] It should be noted that the preset number can be a default value, or can be flexibly set by a user according to actual conditions, and the present embodiment does not make a specific limitation thereon. When the preset number of continuous network layers in the feedforward prediction model are frozen, the preset number of continuous network layers close to the input layer of the feedforward prediction model can be frozen, or other continuous network layers in the feedforward prediction model can be frozen, and the present embodiment does not make a specific limitation thereon.

[0073] In this embodiment, the preset number of continuous network layers in the feedforward prediction model are frozen to retain the basic rules learned by the model, so as to avoid loss of core capabilities due to subsequent model training; then the network layers that are not frozen in the feedforward prediction model are subjected to transfer learning according to the training sample set, so that the model can learn the feedback control strategy in a targeted manner, thereby quickly learning the feedback control strategy without damaging the core capabilities (i.e., the feedforward control strategy). Thus, a comprehensive prediction model coupled with the feedforward control strategy and the feedback control strategy can be formed. It can be understood that the "freezing + iterative training" mode of this embodiment greatly reduces the number of parameters that need to be updated in the model, reduces the training cost, and because the generalization basis of the feedforward prediction model is retained during model training, overfitting of historical error data by the model can be avoided, so as to ensure the prediction stability and prediction accuracy of the comprehensive prediction model obtained by training.

[0074] In another possible implementation, the step S113 can include step S103: Step S103, fine-tuning training is performed on part of the network parameters in the feedforward prediction model according to the training sample set, to generate the comprehensive prediction model.

[0075] It should be noted that, when fine-tuning training is performed on part of the network parameters in the feedforward prediction model according to the training sample set, fine-tuning training can be performed on the parameters (such as scaling factors, offset factors, and weights of features) of the key network layers (such as the output layer, the network layer close to the output, and the normalization layer) in the feedforward prediction model according to the training sample set. By inserting a low-rank matrix into the key network layers in the feedforward prediction model, fine-tuning training can be performed on the parameters of the key network layers.

[0076] In this embodiment, fine-tuning training is performed on part of the network parameters in the feedforward prediction model by using the training sample set, so that the original core prediction capability of the feedforward prediction model is retained, and new error correction rules can be incorporated in a targeted manner. This "partial fine-tuning" approach enables the model to learn the feedback control strategy in a targeted manner, so that the feedback control strategy can be quickly learned without damaging the core capability (i.e., the feedforward control strategy). Thus, a comprehensive prediction model that couples the feedforward control strategy and the feedback control strategy can be formed.

[0077] The above are only two feasible implementation manners of step S113 provided by this embodiment, and this embodiment does not specifically limit the specific implementation manner of step S113.

[0078] Step S12, inputting the current state data and each candidate parameter value of the control parameter into the comprehensive prediction model to obtain a first prediction value of each control target under each candidate parameter value of the control parameter; It should be noted that, in this embodiment, the prediction value of each control target under each candidate parameter value of the control parameter predicted by the comprehensive prediction model is referred to as a first prediction value.

[0079] Step S13, calculating the difference between the demand value of each control target set and the first prediction value of each control target under each candidate parameter value to obtain a control deviation value of each control target under each candidate parameter value.

[0080] In this embodiment, by using a comprehensive prediction model coupled with a feedforward control strategy and a feedback control strategy, the first prediction value of each control target under different candidate parameter values is accurately generated by combining the current state data of the air conditioning equipment and the candidate parameter values of the control parameters. This prediction mechanism not only realizes the early prediction of environmental changes and equipment operation trends through the feedforward control strategy, but also effectively avoids the problems of single feedforward prediction being easily affected by unknown interference or pure feedback prediction having large hysteresis by incorporating the correction logic of historical operation deviation through the feedback control strategy, thereby significantly improving the reliability of the prediction results. On this basis, by subtracting the demand value of each control target from the first prediction value of each control target under each candidate parameter value predicted by the comprehensive prediction model, the control deviation value of each control target under each candidate parameter value can be accurately determined.

[0081] Based on the above-mentioned second embodiment, a third embodiment of the control method of the air conditioning equipment is proposed. In the third embodiment, for the environmental targets in the control targets, the comprehensive prediction model includes a multi-task expert network model; step S12 can include steps S211-S214: Step S211, for any candidate parameter value of the control parameter, inputting the current state data and the candidate parameter value into the multi-task expert network model; It should be noted that the multi-task expert network model (Multigate Mixture of Experts) can simultaneously predict temperature, humidity, and air volume.

[0082] Step S212, based on each expert network in the multi-task expert network model, performing feature extraction processing on the current state data and the candidate parameter value to obtain data processed by each expert network; the expert networks include a temperature expert network, a humidity expert network, an air volume expert network, and a public network; Step S213, based on the gating structure in the multi-task expert network model, using a self-balancing training mechanism to perform fusion processing on the data processed by each expert network to obtain fusion data; It should be noted that the self-balancing training mechanism is a self-adaptive training strategy that dynamically adjusts the weights of components, tasks, or loss terms of a model during the training process of a machine learning or deep learning model, to achieve multi-objective optimization, alleviate data imbalance, or component performance difference problems. The implementation principle of the self-balancing training mechanism can refer to the following formula 2.

[0083] Formula 2; Wherein, Ltotal represents the total loss function (used to measure the deviation between the prediction results of the model and the true results), is a weight coefficient of the i-th expert network, which reflects the importance of each expert network in the final model decision, is a loss function of the i-th expert network (for example, the loss function corresponding to the temperature expert network can measure the gap between the temperature prediction result and the true temperature value) ; is a conditional probability, which represents the probability of the output result being given the model (W is the parameter geometry of the model, and x is the input data) ; and are loss terms related to the model parameters W; and are hyperparameters, similar to variance terms, used to adjust the weight and influence degree of different loss terms.

[0084] Step S214, input the fusion data into each discriminant network in the multi-task expert network model to obtain the temperature prediction value, the humidity prediction value and the air volume prediction value, which are collectively taken as the first prediction value of the environmental target under the candidate parameter value.

[0085] It should be noted that each discriminant network can include a temperature discriminant network, a humidity discriminant network and an air volume discriminant network. The implementation principle of steps S211-S214 in this embodiment can be referred to Figure 3 .

[0086] It can be understood that the existing coupling mechanism limits the simultaneous prediction of the three targets of temperature, humidity and air volume, which is specifically manifested as: when the air conditioning device reduces the temperature, the evaporator surface temperature drops, and the water vapor in the air condenses on the evaporator surface, resulting in a decrease in humidity, so that the change of humidity is subject to the change of temperature; the change of air speed will affect the heat and humidity exchange rate, thereby causing the change of temperature and humidity, for example, high air speed will cause rapid temperature drop at the air outlet, and then speed up the condensation of water vapor to cause a sharp drop in humidity.

[0087] To this end, the embodiment designs a multi-task expert network model including multiple expert networks, a gating structure using a self-balancing training mechanism and multiple discriminant networks. Therefore, by using the multi-task expert network model, the current state data and the candidate parameter value can be combined to simultaneously achieve accurate prediction of temperature, humidity and air volume, thereby accurately determining the first prediction value of the environmental target under the corresponding candidate parameter value, and further improving the control effect and control accuracy of the air conditioning device.

[0088] Based on the above second embodiment, a fourth embodiment of the control method of the air conditioning device is proposed, in which, for the energy saving target in each control target, the comprehensive prediction model comprises a sensible heat ratio prediction model and an energy efficiency prediction model; and step S12 can comprise steps S221-S227: In step S221, for any candidate parameter value of the control parameter, the current state data and the candidate parameter value are input into the input layer of the sensible heat ratio prediction model; In step S222, based on the processing of the feature extraction layer and the temperature and humidity energy distribution layer of the sensible heat ratio prediction model, the first sensible heat ratio of the next operation period and the second sensible heat ratio of the current operation period are obtained. It should be noted that after the processing of the feature extraction layer and the temperature and humidity energy distribution layer of the sensible heat ratio prediction model in this embodiment, the sensible heat ratio of the next operation period and the sensible heat ratio of the current operation period obtained are respectively referred to as the first sensible heat ratio and the second sensible heat ratio.

[0089] In step S223, the first sensible heat ratio and the second sensible heat ratio are subjected to constraint processing in the output layer of the sensible heat ratio prediction model to obtain the predicted sensible heat ratio of the next operation period. It should be noted that when the first sensible heat ratio and the second sensible heat ratio are subjected to constraint processing in the output layer of the sensible heat ratio prediction model, a physical trend loss function can be designed based on the correlation between the sensible heat ratio and the temperature, humidity, air volume and compressor frequency, which can be used to constrain the change direction of the sensible heat ratio of the two periods before and after the constraint; and a sensible heat ratio physical value loss function can be designed based on the energy conservation equation, which can be used to constrain the sensible heat ratio value of the two periods before and after the constraint. On this basis, the output layer of the sensible heat ratio prediction model can use the physical trend loss function and the sensible heat ratio physical value loss function to constrain the first sensible heat ratio and the second sensible heat ratio.

[0090] In addition, it should be noted that the sensible heat ratio prediction model can be a double-channel network, and the implementation principle can refer to Figure 4 . Specifically, after the current state data and the candidate parameter value are input into the input layer of the sensible heat ratio prediction model, the current state data will be used as the input data of the first channel, and the candidate parameter value will be used as the input data of the second channel. After the processing of the feature extraction layer and the temperature and humidity energy distribution layer of the sensible heat ratio prediction model, the first sensible heat ratio SHR i of the next operation period and the second sensible heat ratio SHR i-1 of the current operation period can be obtained; then the first sensible heat ratio SHR i and the second sensible heat ratio SHR i-1 are subjected to constraint processing in the output layer of the sensible heat ratio prediction model, and the predicted sensible heat ratio of the next operation period can be obtained.

[0091] In the case where the sensible heat ratio prediction model is a double-channel network, the implementation process of obtaining the first sensible heat ratio of the next operation period and the second sensible heat ratio of the current operation period based on the processing of the feature extraction layer and the temperature and humidity energy distribution layer of the sensible heat ratio prediction model can include: in the feature extraction layer, performing feature extraction processing on the input data of the first channel and the input data of the second channel to obtain processed first channel data and second channel data; in the temperature and humidity energy distribution layer, determining the first energy consumption caused by temperature change of the next operation period and the second energy consumption caused by humidity change according to the processed first channel data and the second channel data, and determining the first sensible heat ratio of the next operation period and the second sensible heat ratio of the current operation period according to the first energy consumption and the second energy consumption, respectively.

[0092] It should be noted that the next operation period is adjacent to the current operation period, and the energy consumption caused by temperature change and the energy consumption caused by humidity change are approximately the same, so that the sensible heat ratio of the next operation period and the sensible heat ratio of the current operation period can be determined simultaneously by using the energy consumption caused by temperature change and the energy consumption caused by humidity change of the next operation period.

[0093] In addition, it should be noted that the Gaussian elimination algorithm can be used to calculate the first energy consumption caused by temperature change of the next operation period and the second energy consumption caused by humidity change according to the temperature change, humidity change and energy consumption in the processed first channel data, and the temperature change, humidity change and energy consumption in the second channel data.

[0094] Step S224, inputting the current state data and the candidate parameter value into the energy efficiency prediction model; Step S225, obtaining the high-dimensional feature matrix based on the processing of the low-dimensional feature extraction layer and the high-dimensional feature mapping layer of the energy efficiency prediction model; It should be noted that in the processing of the low-dimensional feature extraction layer and the high-dimensional feature mapping layer of the energy efficiency prediction model to obtain the high-dimensional feature matrix, first in the low-dimensional feature extraction layer, the time series data in the current state data and the candidate parameter value can be converted into a graph structure, and then the "node embedding" technology is used to compress the information of each node in the graph structure into a low-dimensional feature vector to obtain each key low-dimensional feature. Then in the high-dimensional feature mapping layer, the key low-dimensional features are aggregated, and the "star operation" mathematical operation is used to map the aggregated features to a high-dimensional nonlinear space to obtain the high-dimensional feature matrix.

[0095] Step S226, performing constraint processing on the high-dimensional feature matrix in the energy efficiency prediction layer of the energy efficiency prediction model to obtain the predicted energy efficiency; It should be noted that, in the energy efficiency prediction layer of the energy efficiency prediction model, when the high-dimensional feature matrix is processed, a physical equation can be designed based on the law of conservation of energy and other physical constraints (for example: input energy = refrigeration capacity + heat loss), and the equation is embedded as a constraint condition in the loss function. Thus, the energy efficiency prediction layer of the energy efficiency prediction model can use the constraint condition embedded in the loss function to process the high-dimensional feature matrix.

[0096] Step S227, the predicted sensible heat ratio and the predicted energy efficiency are taken as the first predicted value of the energy saving target under the candidate parameter value.

[0097] It can be understood that there are two paths for the realization of the energy saving target, namely, sensible heat and energy efficiency. From the perspective of sensible heat, the energy consumed in processing sensible heat is less than that in processing latent heat, and the system with high sensible heat is more energy-saving; from the perspective of energy efficiency, the energy efficiency of the air conditioning equipment refers to the ratio of refrigeration capacity to input power of the air conditioning equipment, the higher the energy efficiency, the higher the refrigeration capacity generated per unit power, and the lower the power required under the same cooling demand, and the equipment is more energy-saving.

[0098] The embodiment accurately determines the predicted sensible heat ratio and the predicted energy efficiency of the air conditioning equipment under the corresponding candidate parameter value by designing the sensible heat ratio prediction model and the energy efficiency prediction model and combining the current state data and the candidate parameter value, thereby accurately determining the first predicted value of the energy saving target under the corresponding candidate parameter value, and further improving the control effect, control adaptability and control accuracy of the air conditioning equipment.

[0099] Based on the above-mentioned second embodiment, the fifth embodiment of the control method of the air conditioning equipment of the present application is proposed, and in the fifth embodiment, for the comfort target in each control target, the comprehensive prediction model includes a comfort parameter prediction model; step S12 can include steps S231-S232: Step S231, for any candidate parameter value of the control parameter, the current state data and the candidate parameter value are input into the comfort parameter prediction model to obtain the human parameter prediction value and the thermal comfort prediction value; wherein the comfort parameter prediction model includes a first model trained with the distribution dispersion degree of the human parameter before and after the running state of the air conditioning equipment is switched as a constraint condition, and a second model trained with the difference of the absolute value of the thermal comfort value before and after the running state of the air conditioning equipment is switched as a constraint condition; It should be noted that the human body parameters of the user can include metabolic rate and clothing thermal resistance (i.e., clothing resistance). The metabolic rate parameter mainly depends on the information of gender, age, weight and activity of the user, among which the activity is a short-term variable feature with a periodical rule, for example, the user exercises from 17:00 to 18:00 every day, and sleeps from 22:00 of the first day to 8:00 of the second day. In addition, the clothing thermal resistance parameter depends on the information of dressing, bedding and living scene, among which the living scene is a key feature with a periodical rule, for example, the user wears clothes and goes out to work from 8:00 to 18:00 every day, wears clothes and does housework from 18:00 to 19:00 every day, and covers the quilt and sleeps from 22:00 of the first day to 8:00 of the second day. Therefore, due to the periodical rule of the activity and the living scene, the user behavior pattern has distribution consistency in the same period, that is, the distribution of the human body parameters of the user before and after the user switches the air conditioning running state has small dispersion in the same period.

[0100] In addition, it should be noted that the thermal comfort value is used to quantitatively evaluate the thermal comfort degree of the human body in a specific thermal environment. It can be understood that the air conditioning operation behavior of the user has a monotonic improvement trend, and when the user controls the air conditioning device to adjust from state A to state B, the thermal comfort of the user should be more comfortable in state B than in state A. The first model can be a PMV model, and the specific construction and workflow thereof can refer to the following: Firstly, a TCN (Temporal Convolutional Network) time series model is constructed by using the correlation coefficient between the state data of the air conditioning device and the working condition parameters of the working condition in which the air conditioning device is located, so as to complete the prediction of the working condition parameters by using the TCN time series model; then, the behavior inertia and behavior optimization criterion of the user are introduced on the basis of the model, so as to construct a double-channel neural network, and the human body parameters are predicted by using the double-channel neural network.

[0101] In addition, it should be noted that the second model can be a UIC model, which can be constructed based on the user adjustment behavior, and the specific construction process can refer to the following: Firstly, a comfort triple (anchor, positive, negative) can be constructed based on the user adjustment behavior, wherein the anchor can be understood as a basic user behavior-thermal comfort state sample for comparison; the positive sample is a sample that is similar to the anchor in thermal comfort experience and belongs to the same comfort feeling; and the negative sample is a sample that is obviously different from the anchor in thermal comfort experience and belongs to different comfort feeling. The comfort triples can then be input into a DASN (Domain Adaption Smoothing Network) to aggregate similar tuples and isolate heterogeneous tuples. DASN processes comfort triples using feature mapping and domain adaptation mechanisms, allowing similar comfort triplets to be more tightly aggregated in the feature space while isolating heterogeneous triplets from each other, thereby clearly distinguishing the distribution of samples with different thermal comfort sensations. Afterwards, the triplet loss function is used to expand the metric distance between positive and negative examples in the feature space through metric learning (positive examples are anchor samples and positive samples, and negative examples are anchor samples and negative samples). That is, the features of similar thermal comfort samples are made more aggregated, and the features of heterogeneous samples are made further apart. By continuously optimizing this distance relationship, the model learns to distinguish different thermal comfort states, and thus a second model can be constructed.

[0102] To help understand the implementation principle of the comfort parameter prediction model, please refer to Figure 5 The architecture diagram of the comfort parameter prediction model shown in the figure can include an input layer, a feature extraction layer, a data conversion layer, a PMV parameter layer, and an output layer. The input layer is located in the first layer of the comfort parameter prediction model. During the model training process, the input layer is used to receive model output parameters. The model output parameters include the historical state data (History periods Data) of the air conditioning equipment, the state data of the air conditioning equipment in state A (A State Current Data), and the state data of the air conditioning equipment in state B (B State Current Data). State A represents the air conditioning operating state before the air conditioning equipment state is switched based on a user instruction within a first preset time period. State B represents the air conditioning operating state after the air conditioning equipment state is switched based on the user instruction within the first preset time period. In addition, each set of state features includes 9 features, namely return air temperature, condenser temperature, evaporator temperature, outdoor temperature, exhaust temperature, return air humidity, internal fan speed, compressor frequency, and temperature difference. The feature extraction layer uses an MLP (Multi-Layer Perceptron) network, represented as MLPExtracting Layer, which is used to extract features from the air conditioning state data before and after the state switching of the air conditioning equipment, and output human body parameters, including the metabolic rate under state A. Clothes resistance , and the metabolic rate in state B Clothes resistance ; the data conversion layer is represented as a Frozen CTCN Model, and is used to determine thermal environment parameters based on historical state data, state data of the air conditioner in state A, and state data of the air conditioner in state B, the thermal environment parameters including air temperature , relative humidity , convective heat transfer coefficient , and mean radiant temperature in state A, and air temperature , relative humidity , convective heat transfer coefficient , and mean radiant temperature in state B; the PMV parameter layer is used to align the human body parameters and the thermal environment parameters in state A and state B respectively, as inputs of the output layer; and the output layer is used to calculate the thermal comfort values in state A and state B based on a calculation formula of the PMV value and . That is, it can be understood that the first model and the second model can be used as a whole for model training and model application, and the overall comfort parameter prediction model has the ability to predict human body parameters and thermal comfort values.

[0103] In step S232, the human body parameter prediction value and the thermal comfort prediction value are used as the first prediction value of the comfort target under the candidate parameter value.

[0104] The embodiment designs a comfort parameter prediction model, which includes a first model for predicting human body parameters and a second model for predicting thermal comfort. Thus, by using the comfort parameter prediction model, the human body parameter prediction value and the thermal comfort prediction value can be accurately predicted by combining the current state data and the candidate parameter value; then, the human body parameter prediction value reflecting the thermal comfort from the human physiological aspect and the thermal comfort prediction value reflecting the thermal comfort from the user behavior aspect are used as the first prediction value of the comfort target under the candidate parameter value, so that the condition that the comfort target can achieve under the corresponding candidate parameter value can be accurately analyzed, thereby realizing diversified and differentiated comfort evaluation, and thus being beneficial to further improving the control effect and control accuracy of the air conditioning equipment.

[0105] Based on the first embodiment, a sixth embodiment of the control method of the air conditioning equipment is provided, in which step S10 can include steps S14-S19: In step S14, current state data of the air conditioning equipment is obtained. In step S15, the current state data and each candidate parameter value of the control parameter are input into a feedforward prediction model associated with a feedforward control strategy, to obtain a second prediction value of each control target under each candidate parameter value. It should be noted that the predicted value of each control target at each candidate parameter value predicted by the feedforward control strategy is referred to as a second predicted value in this embodiment. The feedforward prediction model is a model trained in advance before the air conditioning device is deployed, and the feedback prediction model is a model trained according to the actual operation of the air conditioning device after the air conditioning device is deployed.

[0106] In step S16, a difference between the demand value of each control target set and the second predicted value of each control target at each candidate parameter value is calculated to obtain a first deviation value of each control target at each candidate parameter value. In step S17, a third predicted value of each control target at the current parameter value of the control parameter predicted by the feedback prediction model associated with the feedback control strategy in the previous operation period is obtained. It should be noted that the predicted value of each control target at the current parameter value of the control parameter predicted by the feedback prediction model is referred to as a third predicted value in this embodiment.

[0107] In a feasible implementation, the construction step of the feedback prediction model can include steps S171-S173. In step S171, a difference between the demand value and the actual value of each control target in a plurality of historical operation periods before the current operation period is obtained to obtain a correction loss of each control target in the plurality of historical operation periods. In step S172, a training sample set is constructed according to the correction loss of each control target in the plurality of historical operation periods. In step S173, reinforcement learning is performed on the to-be-trained model according to the training sample set to generate the feedback prediction model.

[0108] It should be noted that when reinforcement learning is performed on the to-be-trained model according to the training sample set, the training sample set can be divided into a training set and a verification set according to a certain proportion. Thus, reinforcement learning is performed on the to-be-trained model using the training set, and whether the model after reinforcement learning converges is verified using the verification set. After the model after reinforcement learning converges, the feedback prediction model is generated.

[0109] It should be noted that the to-be-trained model can be a neural network, a linear regression model, etc., and this embodiment does not make a specific limitation.

[0110] In this implementation, after the training sample set is constructed using the difference between the demand value and the actual value of each control target in a plurality of historical operation periods before the current operation period, i.e., the correction loss of each control target in the plurality of historical operation periods, reinforcement learning is directly performed on the to-be-trained model using the training sample set to generate the feedback prediction model.

[0111] Step S18, calculate the difference between the actual value of each control target in the current operation period and the third predicted value of each control target, to obtain the second deviation value of each control target; Step S19, the first deviation value and the second deviation value of each control target under each candidate parameter value are weighted and fused to obtain the control deviation value of each control target under each candidate parameter value.

[0112] It should be noted that in the process of weighting and fusing the first deviation value and the second deviation value of each control target under each candidate parameter value, the weight value of each control target under the feedforward control strategy and the feedback control strategy can be a default value, or can be flexibly determined according to the actual situation (for example, the weight values of the feedforward control strategy and the feedback control strategy will change accordingly under different operation functions of the air conditioning equipment), and the embodiment does not make specific limitation.

[0113] In this embodiment, by respectively using the feedforward prediction model associated with the feedforward control strategy and the feedback prediction model associated with the feedback control strategy, the advance prediction of environmental changes and equipment operation trends is realized by means of the feedforward control strategy, and the correction logic of historical operation deviation is introduced by means of the feedback control strategy, thereby effectively avoiding the problems of single feedforward prediction being easily affected by unknown interference or pure feedback prediction having large hysteresis, and significantly improving the reliability of the prediction result. On this basis, by weighting and fusing the deviation value determined by means of the feedforward control strategy and the deviation value determined by means of the feedback control strategy, the control deviation value of each control target under each candidate parameter value can be accurately determined.

[0114] Based on the second embodiment, the third embodiment, the fourth embodiment, the fifth embodiment and / or the sixth embodiment, a seventh embodiment of the control method of the air conditioning equipment is proposed, and in the seventh embodiment, the step of obtaining the current state data of the air conditioning equipment can include steps S01-S04: Step S01, obtaining the current energy of the compressor in the current operation period of the air conditioning equipment and the historical energy of the compressor in a plurality of historical operation periods before the current operation period; It should be noted that the current energy of the compressor is the energy of the compressor in the current operation period, and the historical energy of the compressor is the energy of the compressor in the historical operation period. The energy can be the frequency, refrigerating capacity, heating capacity, energy consumption, power, etc. of the compressor, and the embodiment does not make specific limitation. The number of historical energy of the compressor obtained can be set according to the number of operation periods with the most energy lag.

[0115] Step S02, obtaining the inertial energy of each historical operation period transferred to the current operation period according to the historical energy of the compressor in the plurality of historical operation periods; It should be noted that the inertial energy refers to the energy of the compressor in the historical operation period affected by the energy transmission inertia, which is transmitted to the current operation period.

[0116] In a possible implementation, the step S02 can include steps S021-S022. In the step S021, the energy transmission inertia coefficient of the compressor historical energy in each historical operation period transmitted to the current operation period is obtained. It should be noted that the energy transmission inertia coefficient is used to quantify the hysteresis effect existing in the energy transmission process.

[0117] In a possible implementation, the step S021 can include: calculating the time interval between the current operation period and each historical operation period; and determining the energy transmission inertia coefficient of the compressor historical energy in each historical operation period transmitted to the current operation period according to the time interval between the current operation period and each historical operation period.

[0118] The embodiment is not limited to the specific implementation of the step S021. For example, in other possible implementations, the time interval and the operation parameter (such as the compressor frequency, the expansion valve angle, etc.) corresponding to each historical operation period can be considered at the same time to determine the energy transmission inertia coefficient of the compressor historical energy in each historical operation period transmitted to the current operation period (i.e., one time interval and one operation parameter correspond to one energy transmission inertia coefficient), so as to further ensure the accuracy of the energy transmission inertia coefficient corresponding to each historical operation period, thereby further improving the control accuracy of the air conditioning equipment.

[0119] In the determination of the energy transmission inertia coefficient of the compressor historical energy in each historical operation period transmitted to the current operation period according to the time interval between the current operation period and each historical operation period, the time interval between the current operation period and each historical operation period can be respectively indexed to find the energy transmission inertia coefficient corresponding to each time interval in the preset relationship table, as the energy transmission inertia coefficient of the compressor historical energy in each historical operation period transmitted to the current operation period. The time interval between the current operation period and each historical operation period can also be respectively input into a pre-trained energy transmission inertia model to obtain the energy transmission inertia coefficient of the compressor historical energy in each historical operation period transmitted to the current operation period. The embodiment is not limited to this.

[0120] It should be noted that the relationship table is used to record the mapping relationship between the time interval and the energy transfer inertia coefficient. The enthalpy difference measurement method or expert experience can be used to construct the relationship table. When the enthalpy difference measurement method is used to construct the relationship table, the compressor frequency, fan speed, expansion valve angle and other parameters can be fixed in the static enthalpy difference experiment measurement, and then the indoor reaches a stable state through the load generator. The energy transfer inertia disappears in the stable state (i.e. the energy accumulated from the historical running period to the current running period is approximately equal to the energy transferred from the current running period to the future running period). On the basis of the experiment, the condition and the load remain unchanged, and the detection of the increase in refrigerating capacity and time is carried out when the compressor frequency parameter is increased. Through the control variable method, the energy transfer inertia coefficient can be estimated based on the incremental difference experiment of multiple rounds.

[0121] In addition, it should be noted that the energy transfer inertia model is used to predict the energy transfer inertia coefficient, which can be constructed by using the numerical driving modeling method. Specifically, in the dynamic environment experiment measurement, the environment state and the running state of the current running period and multiple historical running periods are taken as independent variables, the energy transferred from the current running period to the future W periods is taken as the dependent variable, and the refrigerating capacity obtained by the enthalpy difference experiment is taken as the calibration. The energy transfer inertia model is trained to minimize the error between the accumulated refrigerating capacity of the current running period and the refrigerating capacity obtained by the enthalpy difference experiment (i.e. the loss function constructed converges).

[0122] In step S022, the inertial energy transferred from each historical running period to the current running period is calculated according to the compressor historical energy of each historical running period and the energy transfer inertia coefficient of the compressor historical energy of each historical running period to the current running period.

[0123] It should be noted that the inertial energy transferred from each historical running period to the current running period can be obtained by multiplying the compressor historical energy of each historical running period and the energy transfer inertia coefficient thereof.

[0124] In step S03, the accumulated energy of the current running period is obtained according to the inertial energy and the current energy of the compressor. It should be noted that the accumulated energy is the energy accumulated in the current running period by the inertial energy and the current energy of the compressor.

[0125] In a feasible implementation, step S03 can include calculating the sum of the inertial energy and the current energy of the compressor to obtain the accumulated energy of the current running period.

[0126] In another possible implementation, step S03 can include: obtaining the energy transfer inertia coefficient corresponding to the current operation period; calculating the product of the current energy of the compressor and the energy transfer inertia coefficient corresponding to the current operation period, to obtain the target energy of the current energy transfer of the compressor in the current operation period; and calculating the sum of the inertia energies and the target energy, to obtain the accumulated energy of the current operation period.

[0127] It can be understood that, considering that the energy transfer inertia is essentially a hysteresis effect of energy transfer, thus, the current energy of the compressor in the current operation period is affected by the energy transfer inertia, and it can also not be fully applied in the current operation period. Therefore, in the process of obtaining the accumulated energy of the current operation period according to the inertia energies and the current energy of the compressor, the embodiment first determines the target energy of the current energy transfer of the compressor in the current operation period by using the energy transfer inertia coefficient corresponding to the current operation period; and then integrates the inertia energies and the target energy to determine the accumulated energy of the current operation period. Therefore, the embodiment not only considers the influence of the historical energy of the compressor in the historical operation period on the current operation period under the energy transfer inertia, but also considers the influence of the current energy of the compressor in the current operation period on the current operation period under the energy transfer inertia, to more effectively solve the influence of the energy transfer inertia on the control accuracy of the air conditioning device, thereby further improving the control accuracy of the air conditioning device.

[0128] The above are only two possible implementations of step S03 provided by the embodiment, and the embodiment does not specifically limit the specific implementation of step S03.

[0129] In addition, in combination with the implementation of step S03 and step S022 of the embodiment, the obtained implementation process can be represented as formula 3 below, and the implementation principle represented by formula 3 can be specifically referred to Figure 6 .

[0130] Formula 3; Wherein, is the accumulated energy of the current operation period, is the energy transfer coefficient of the jth operation period to the ith operation period (i.e., the current operation period), is the compressor energy of the jth operation period, and W is the number of operation periods in which energy is most likely to lag (the number of operation periods in which energy is most likely to lag includes one current operation period and multiple historical operation periods, which can be a default value, for example, 10, which can include one current operation period and nine historical operation periods, or can be flexibly set by the user according to the actual situation, and the embodiment does not specifically limit this.

[0131] In step S04, the current state data of the air conditioning device is determined based on the accumulated energy.

[0132] In an implementation, step S04 can include step S041: In step S041, the state data corresponding to the accumulated energy is obtained as the current state data based on a preset mapping relationship between the compressor energy and the state data.

[0133] It should be noted that the mapping relationship between the compressor energy and the state data can be recorded in the form of a relationship table, a key-value pair, etc., and the present embodiment does not make a specific limitation thereon.

[0134] It can be known from the above that, in the determination of the current state data of the air conditioning device, the present embodiment not only considers the current energy of the compressor in the current running period, but also considers the historical energy of the compressor in the plurality of historical running periods before the current running period, which is respectively transferred to the inertial energy in the current running period, so as to effectively solve the influence of the energy transfer inertia on the control accuracy of the air conditioning device, thereby accurately determining the actual energy situation of the compressor of the air conditioning device in the current running period, and further accurately determining the current state data of the air conditioning device, thereby improving the control accuracy of the air conditioning device.

[0135] The present embodiment also provides an air conditioning device, which includes at least one processor, and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the control method of the air conditioning device in the above embodiments.

[0136] Reference will be made to the following Figure 7 which shows a structural schematic diagram of an air conditioning device suitable for implementing the present embodiment. Figure 7 The air conditioning device shown is merely an example, and should not bring any limitation to the function and use range of the present embodiment.

[0137] As Figure 7As shown, the air conditioning device can include a processing device 101 (e.g., a central processing unit, a graphics processing unit, etc.) that can perform various appropriate actions and processes according to programs stored in a read-only memory 102 or loaded from a storage device 103 into a random access memory 104. Various programs and data required for operation of the air conditioning device are also stored in the random access memory 104. The processing device 101, the read-only memory 102, and the random access memory 104 are connected to each other by a bus 105. An input / output interface 106 is also connected to the bus 105. Generally, the following systems can be connected to the input / output interface 106: input devices 107 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 108 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; the storage device 103 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 109. The communication device 109 can allow the air conditioning device to communicate wirelessly or wired with other devices to exchange data. Although the air conditioning device with various systems is shown in the figure, it should be understood that all of the shown systems are not required to be implemented or possessed. More or less systems can be alternatively implemented or possessed.

[0138] In particular, the processes described above with reference to the flowcharts can be implemented as a computer software program according to embodiments of the present disclosure. For example, embodiments of the present disclosure include a computer program product comprising a computer program carried on a computer readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network through the communication device, or installed from the storage device 103, or installed from the read-only memory 102. When the computer program is executed by the processing device 101, the above-mentioned functions defined in the methods of the embodiments of the present disclosure are performed.

[0139] The air conditioning device provided by the embodiments of the present disclosure adopts the control method of the air conditioning device in the above-mentioned embodiments, and can simultaneously improve the control effect and control adaptability of the air conditioning device. Compared with the prior art, the air conditioning device provided by the embodiments of the present disclosure has the same beneficial effects as the control method of the air conditioning device provided by the above-mentioned embodiments, and other technical features in the air conditioning device are the same as the features disclosed in the above-mentioned embodiments, which will not be described here.

[0140] It should be understood that parts of the embodiments of the present disclosure can be realized by hardware, software, firmware or a combination thereof. In the description of the above-mentioned embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0141] The above merely provides the specific implementation of the embodiments of the present application, but the protection scope of the embodiments of the present application is not limited thereto, any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the embodiments of the present application. Therefore, the protection scope of the embodiments of the present application should be subject to the protection scope of the above claims.

[0142] The embodiments of the present application further provide a computer readable storage medium, which stores a computer program capable of running on a processor, and the computer program is used for executing the control method of the air conditioning device in the above embodiments.

[0143] The computer readable storage medium provided by the embodiments of the present application may, for example, be a U disk, but is not limited to an electric, magnetic, optical, electromagnetic, infrared, or semiconductor system or device, or any combination of the above. More specific examples of the computer readable storage medium can include, but are not limited to, an electric connection with one or more conductive lines, a portable computer disk, a hard disk, a random access memory (RAM), a read only memory (ROM), an erasable programmable read only memory (EPROM or flash memory), an optical fiber, a portable compact disk read only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the embodiments, the computer readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, a system or a device. The program code contained in the computer readable storage medium can be transmitted by any suitable medium, including but not limited to an electric wire, an optical cable, an RF (Radio Frequency), etc., or any suitable combination of the above.

[0144] The above computer readable storage medium can be contained in the air conditioning device, or can exist separately without being assembled into the air conditioning device.

[0145] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the air conditioning equipment, the air conditioning equipment is caused to: obtain multiple candidate parameter values ​​of the control parameters of the air conditioning equipment in the next operating cycle, and based on the feedforward control strategy and the feedback control strategy, predict the control deviation value of each control target of the air conditioning equipment under each candidate parameter value of the control parameter; the control target is used to characterize the performance index that the air conditioning equipment needs to achieve, and the control deviation value is used to characterize the extent to which the control target meets the set control target requirement value; obtain the target weight value of each control target under the current control requirement of the air conditioning equipment; based on the target weight value of each control target, the control deviation value of each control target under each candidate parameter value is weightedly fused to obtain the comprehensive deviation value of each candidate parameter value; the candidate parameter value with the smallest comprehensive deviation value among the candidate parameter values ​​is used as the target parameter value, and the air conditioning equipment is controlled to operate according to the target parameter value of the control parameter in the next operating cycle.

[0146] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0147] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0148] The modules involved in the embodiments of the present application can be implemented in software or hardware. In some cases, the names of the modules do not limit the modules themselves.

[0149] The computer readable storage medium provided by the embodiments of the present application stores computer readable program instructions for executing the control method of the air conditioning device, and can improve the control effect and control adaptability of the air conditioning device at the same time. Compared with the prior art, the beneficial effects of the computer readable storage medium provided by the embodiments of the present application are the same as those of the control method of the air conditioning device provided by the above embodiments, and are not repeated here.

[0150] The embodiments of the present application also provide a computer program product, comprising a computer program, and the computer program is executed by a processor to implement the steps of the control method of the air conditioning device as described above.

[0151] The computer program product provided by the embodiments of the present application can improve the control effect and control adaptability of the air conditioning device at the same time. Compared with the prior art, the beneficial effects of the computer program product provided by the embodiments of the present application are the same as those of the control method of the air conditioning device provided by the above embodiments, and are not repeated here.

[0152] The above is only the preferred embodiments of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation using the content of the specification and drawings of the present application, or direct or indirect application in other related technical fields, are also included in the patent processing scope of the present application.

Claims

1. A method for controlling an air conditioning device, characterized in that: The method comprises: Acquiring multiple candidate parameter values ​​for a control parameter of the air conditioning equipment in the next operating cycle, and predicting, based on a feedforward control strategy and a feedback control strategy, a control deviation value for each control target of the air conditioning equipment at each candidate parameter value of the control parameter; the control target is used to represent a performance indicator to be achieved by the air conditioning equipment, and the control deviation value is used to represent the degree to which the control target meets a set control target requirement value; Obtaining a target weight value of each control target under the current control demand of the air conditioning equipment; Based on the target weight value of each control target, a weighted fusion process is performed on the control deviation value of each control target under each candidate parameter value to obtain a comprehensive deviation value of each candidate parameter value; The candidate parameter value with the smallest comprehensive deviation value among the candidate parameter values ​​is used as the target parameter value, and the air conditioning equipment is controlled to operate according to the target parameter value of the control parameter in the next operation cycle.

2. The method according to claim 1, wherein The step of predicting the control deviation value of each control target of the air conditioning equipment under each candidate parameter value of the control parameter based on the feedforward control strategy and the feedback control strategy includes: Obtaining a comprehensive prediction model that couples a feedforward control strategy and a feedback control strategy, and obtaining current state data of the air conditioning equipment; Inputting the current state data and each candidate parameter value of the control parameter into the comprehensive prediction model to obtain a first prediction value of each control target under each candidate parameter value of the control parameter; The difference between the set demand value of each of the control targets and the first predicted value of each of the control targets under each of the candidate parameter values ​​is calculated to obtain the control deviation value of each of the control targets under each of the candidate parameter values.

3. The method according to claim 2, wherein The step of obtaining a comprehensive prediction model that couples the feedforward control strategy and the feedback control strategy includes: Obtaining the difference between the demand value and the actual value of each control target in multiple historical operation cycles before the current operation cycle, and obtaining the correction loss of each control target in the multiple historical operation cycles; Constructing a training sample set based on the correction loss of each of the control objectives in multiple historical operation cycles; The pre-trained feedforward prediction model is retrained based on the training sample set to generate the comprehensive prediction model.

4. The method according to claim 3, wherein The step of retraining the pre-trained feedforward prediction model based on the training sample set to generate the comprehensive prediction model includes: Freezing a preset number of consecutive network layers in the feedforward prediction model; Based on the training sample set, transfer learning is performed on the unfrozen network layers in the feedforward prediction model to generate the comprehensive prediction model.

5. The method according to claim 3, wherein The step of retraining the pre-trained feedforward prediction model based on the training sample set to generate the comprehensive prediction model further includes: Based on the training sample set, some network parameters in the feedforward prediction model are fine-tuned to generate the comprehensive prediction model.

6. The method according to claim 2, wherein For the environmental target in each of the control targets, the comprehensive prediction model includes a multi-task expert network model; The step of inputting the current state data and each candidate parameter value of the control parameter into the comprehensive prediction model to obtain a first predicted value of each control target under each candidate parameter value of the control parameter comprises: For any candidate parameter value of the control parameter, inputting the current state data and the candidate parameter value into the multi-task expert network model; Based on each expert network in the multi-task expert network model, feature extraction processing is performed on the current state data and the candidate parameter values ​​to obtain data processed by each expert network; each expert network includes a temperature expert network, a humidity expert network, an air volume expert network and a public network; Based on the gate control structure in the multi-task expert network model, a self-balancing training mechanism is adopted to fuse the data processed by each expert network to obtain fused data; The fused data is input into each discriminant network in the multi-task expert network model to obtain a temperature prediction value, a humidity prediction value and an air volume prediction value, which are collectively used as the first prediction value of the environmental target under the candidate parameter value.

7. The method according to claim 2, wherein For the energy-saving target in each of the control targets, the comprehensive prediction model includes a sensible heat ratio prediction model and an energy efficiency prediction model; The step of inputting the current state data and each candidate parameter value of the control parameter into the comprehensive prediction model to obtain a first predicted value of each control target under each candidate parameter value of the control parameter comprises: For any candidate parameter value of the control parameter, inputting the current state data and the candidate parameter value into an input layer of the sensible heat ratio prediction model; Based on the processing of the feature extraction layer and the temperature and humidity energy distribution layer of the sensible heat ratio prediction model, a first sensible heat ratio of the next operation cycle and a second sensible heat ratio of the current operation cycle are obtained; At the output layer of the sensible heat ratio prediction model, constraining the first sensible heat ratio and the second sensible heat ratio to obtain a predicted sensible heat ratio for the next operation cycle; and, inputting the current state data and the candidate parameter values ​​into the energy efficiency prediction model; Based on the processing of the low-dimensional feature extraction layer and the high-dimensional feature mapping layer of the energy efficiency prediction model, a high-dimensional feature matrix is ​​obtained; At the energy efficiency prediction layer of the energy efficiency prediction model, constraining the high-dimensional feature matrix to obtain predicted energy efficiency; The predicted sensible heat ratio and the predicted energy efficiency are used together as the first predicted value of the energy-saving target under the candidate parameter value.

8. The method according to claim 2, wherein For the comfort target in each of the control targets, the comprehensive prediction model includes a comfort parameter prediction model; The step of inputting the current state data and each candidate parameter value of the control parameter into the comprehensive prediction model to obtain a first predicted value of each control target under each candidate parameter value of the control parameter comprises: For any candidate parameter value of the control parameter, the current state data and the candidate parameter value are input into the comfort parameter prediction model to obtain a human body parameter prediction value and a thermal comfort prediction value; wherein the comfort parameter prediction model includes a first model trained using the distribution dispersion of the human body parameters before and after the air conditioning equipment operating state is switched as a constraint condition, and a second model trained using the difference in the absolute value of the thermal comfort value before and after the air conditioning equipment operating state is switched as a constraint condition; The human body parameter prediction value and the thermal comfort prediction value are used together as the first prediction value of the comfort target under the candidate parameter value.

9. The method according to claim 1, wherein The step of predicting the control deviation value of each control target of the air conditioning equipment under each candidate parameter value of the control parameter based on the feedforward control strategy and the feedback control strategy further includes: Acquiring current status data of the air conditioning equipment; Inputting the current state data and each candidate parameter value of the control parameter into a feedforward prediction model associated with a feedforward control strategy to obtain a second predicted value of each control target under each candidate parameter value; Calculating the difference between the set demand value of each of the control targets and the second predicted value of each of the control targets under each of the candidate parameter values ​​to obtain a first deviation value of each of the control targets under each of the candidate parameter values; Obtaining a third predicted value of each of the control targets under the current parameter value of the control parameter predicted by a feedback prediction model associated with the feedback control strategy in a previous operation cycle; Calculating the difference between the actual value of each control target in the current operation cycle and the third predicted value of each control target to obtain a second deviation value of each control target; A weighted fusion process is performed on the first deviation value and the second deviation value of each control target under each candidate parameter value to obtain a control deviation value of each control target under each candidate parameter value.

10. The method according to claim 9, wherein The steps of constructing the feedback prediction model include: Obtaining the difference between the demand value and the actual value of each control target in multiple historical operation cycles before the current operation cycle, and obtaining the correction loss of each control target in the multiple historical operation cycles; Constructing a training sample set based on the correction loss of each of the control objectives in multiple historical operation cycles; Based on the training sample set, reinforcement learning is performed on the training model to generate the feedback prediction model.

11. The method according to any one of claims 2 to 10, characterized in that The step of obtaining the current status data of the air conditioning equipment includes: Acquire the current energy of the compressor of the current operation cycle of the air conditioning equipment and the historical energy of the compressor of multiple historical operation cycles before the current operation cycle; Obtaining inertial energy transferred from each historical operating cycle to the current operating cycle based on historical compressor energy of the plurality of historical operating cycles; Obtaining cumulative energy of a current operation cycle according to each of the inertial energies and the current energy of the compressor; Based on the accumulated energy, current status data of the air conditioning device is determined.

12. The method according to claim 11, wherein The step of obtaining the inertial energy transferred from each historical operating cycle to the current operating cycle based on the compressor historical energy of the multiple historical operating cycles includes: Obtaining the energy transfer inertia coefficient of the compressor from historical energy of each historical operating cycle to the current operating cycle; The inertial energy transferred from each historical operating cycle to the current operating cycle is calculated based on the compressor historical energy of each historical operating cycle and the energy transfer inertia coefficient transferred from the compressor historical energy of each historical operating cycle to the current operating cycle.

13. The method according to claim 11, wherein The step of determining the current state data of the air conditioning equipment based on the accumulated energy comprises: Based on a preset mapping relationship between compressor energy and state data, state data corresponding to the accumulated energy is acquired as the current state data.

14. The method according to any one of claims 1 to 10, characterized in that Each of the control targets includes an environmental target, a comfort target, and an energy-saving target. The step of obtaining a target weight value of each of the control targets under the current control requirements of the air conditioning equipment includes: Calculating the time interval between the current operation cycle and the moment when the user last set each of the control target demand values, and obtaining the current electricity price; Determining an initial weight value of the environmental target and an initial weight value of the comfort target according to the time interval; the initial weight value of the environmental target is negatively correlated with the time interval, and the initial weight value of the comfort target is positively correlated with the time interval; determining an initial weight value of the energy-saving target according to the electricity price; the initial weight value of the energy-saving target is positively correlated with the electricity price; The initial weight value of the environmental target, the initial weight value of the comfort target, and the initial weight value of the energy-saving target are normalized to obtain the target weight value of the environmental target, the target weight value of the comfort target, and the target weight value of the energy-saving target.

15. The method according to any one of claims 1 to 10, characterized in that The step of obtaining a target weight value of each control target under the current control demand of the air conditioning equipment further includes: Based on a mapping relationship between a preset operating function and a weight value of each control target, the weight value of each control target corresponding to the current operating function of the air conditioning equipment is obtained as a target weight value of each control target.

16. The method according to any one of claims 1 to 10, characterized in that The step of obtaining multiple candidate parameter values ​​of the control parameters of the air conditioning equipment includes: Obtaining historical parameter values ​​of the control parameters in the previous operation cycle; A plurality of candidate parameter values ​​of the control parameter are determined based on the historical parameter values.

17. An air conditioning device, characterized in that: The air conditioning device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the air conditioning device control method according to any one of claims 1 to 16.

18. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for controlling an air conditioning device according to any one of claims 1 to 16.

Images