Air conditioning system and control method of unit of air conditioning system

Abstract

The invention provides a control method for an air conditioning unit. The control method comprises the steps that operation of the air conditioning unit is controlled; acquiring operation data of the air conditioning unit in real time; selectively performing auxiliary control on the operation of the air conditioning unit based on the acquired operation data of the air conditioning unit; when auxiliary control is carried out on operation of the air conditioning unit, recommended input control data is generated based on the determined air conditioning unit model; and operation of the air conditioning unit is controlled based on the recommended input control data. The air conditioning unit model is created by inputting the convex neural network, so that the model has better prediction smoothness, reliability and robustness. The optimal control data is searched in the adjacent interval of the current control parameter as the recommended control data, so that the search density can be increased, the extreme value search speed can be increased, and the operation safety problem can be prevented. Seamless switching between original control and auxiliary control of the air conditioning unit can be achieved, so that the air conditioning unit is more energy-saving in operation, longer in service life and the like.

Description

Technical Field

[0001] The present application relates to an air conditioning system, and in particular to a control method for a unit in the air conditioning system. Background Art

[0002] In modern refrigeration technology, chillers, such as centrifugal chillers, are widely used in large commercial buildings, industrial refrigeration, medical treatment, food processing, data centers, and AI computing centers due to their high efficiency, stability, and reliability. Taking the data centers that have been booming in recent years as an example, the energy consumption of ventilation and air-conditioning systems often accounts for a large part of the total energy consumption. Therefore, improving efficiency and reducing energy consumption have always been important issues in the air-conditioning and refrigeration industry. Summary of the invention

[0003] The inventors found that energy-saving technologies for large cooling systems are mainly divided into two categories. One is the optimization control at the system operation level. Reasonable setting parameters are set for each component, while coordinating load requirements. System-level control often needs to provide targeted solutions based on actual site conditions due to factors such as on-site load and system design differences. Second, optimization control at the unit equipment level, that is, through a better control strategy, the coefficient of performance (COP) of the air-conditioning equipment is made higher.

[0004] The present application relates to optimization control at the unit equipment level. A method for controlling the operation of an air-conditioning unit of an air-conditioning system is to obtain recommended control data based on an air-conditioning unit model to operate the air-conditioning unit under the control data. The inventors have found that modeling of air-conditioning units based on pure data models requires a large amount of test data. If the data sample is insufficient, the following defects will exist: 1) The prediction model is not robust enough, and excessive prediction deviations will occur in areas where training data is lacking, resulting in erroneous control actions based on the control output predicted by the model; 2) Compared with the actual physical response, the predicted output of the model under monotonic input is not smooth, and there is often large fluctuation noise, which makes the subsequent optimization efficiency worse or even impossible to find the optimal solution.

[0005] In order to solve the above problems, the present application creates an air-conditioning unit model through a convex neural network, and searches for better control data based on the created air-conditioning unit model, so that the air-conditioning unit operates under better control data. The air-conditioning unit model created by the convex neural network has better prediction smoothness, reliability and robustness. In addition, when the present application searches for better control data based on the created air-conditioning unit model, it searches for better control data in the vicinity of the current control data of the unit, which can narrow the search range, increase the search density, increase the speed of extreme value search, and search for results closer to the extreme value, and can prevent the deviation between the searched optimized control data and the current control data from being too large to cause operational safety problems. The present application also provides an auxiliary control method combined with the original control of the air-conditioning system to achieve seamless switching between auxiliary control and original control.

[0006] Specifically, according to the first aspect of the present application, the present application provides a control method for an air-conditioning unit. The control method comprises steps S1-S5. In step S1, the operation of the air-conditioning unit is controlled. In step S2, the operation data of the air-conditioning unit is acquired in real time. In step S3, based on the acquired operation data of the air-conditioning unit, the operation of the air-conditioning unit is selectively auxiliary controlled. In step S4, when the operation of the air-conditioning unit is auxiliary controlled, recommended input control data is generated based on a determined air-conditioning unit model. In step S5, the operation of the air-conditioning unit is controlled based on the recommended input control data.

[0007] According to the first aspect of the present application, in step S5: when the recommended input control data is the same as the current operating data of the air-conditioning unit, maintain the original control of the operation of the air-conditioning unit and go to step S1; when the recommended input control data is different from the current operating data of the air-conditioning unit, control the operation of the air-conditioning unit based on the recommended updated input control data.

[0008] According to the first aspect of the present application, the control method further includes steps S6 and S7. In step S6, when the operation of the air-conditioning unit is controlled based on the recommended updated input control data, the updated operation data of the air-conditioning unit is obtained to determine whether it meets the operation conditions. In step S7, if the updated operation data all meet the operation conditions, the next control cycle is performed and the control is transferred to step S1, otherwise, the control is switched to the original control of the operation of the air-conditioning unit.

[0009] According to the first aspect of the present application, in step S4, recommended input control data is generated by running an optimizer. Running the optimizer includes steps S4.1-S4.4. In step S4.1, one or more actual input data corresponding to at least a part of the input parameters of the air-conditioning unit model in the current operation data of the air-conditioning unit is obtained; in step S4.2, an optimized search range of the corresponding input parameters of the air-conditioning unit model is generated based on the one or more actual input data obtained. In step S4.3, multiple input data in the optimized search range are obtained, and the obtained multiple input data are input into the air-conditioning unit model to generate multiple predicted output data. In step S4.4, the optimized predicted output data and the optimized input data corresponding to the optimized predicted output data are obtained from the multiple predicted output data, and the optimized input data is used as the recommended input control data.

[0010] According to the first aspect of the present application, in step S4.4, obtaining optimized predicted output data and optimized input data corresponding to the optimized predicted output data includes: comparing multiple predicted output data with corresponding actual output data to screen out predicted output data that is smaller than the corresponding actual output data; removing a portion of the predicted output data and the input data corresponding to the predicted output data according to safe operating rules; and obtaining the optimal predicted output data and the optimized input data corresponding to the optimal predicted output data from the predicted output data after screening and removal.

[0011] According to the first aspect of the present application, in step S4.4, filtering out predicted output data that is smaller than the corresponding actual output data includes: filtering out predicted input power that is smaller than the actual input power of the air-conditioning unit; obtaining the optimal predicted output data includes: obtaining the optimal predicted input power of the air-conditioning unit.

[0012] According to the first aspect of the present application, in step S3, selectively performing auxiliary control on the operation of the air-conditioning unit includes: judging whether the operation of the air-conditioning unit meets the stability requirements based on the acquired operation data of the air-conditioning unit; when the operation of the air-conditioning unit meets the stability requirements, performing auxiliary control on the operation of the air-conditioning unit to generate recommended input control data; when the operation of the air-conditioning unit does not meet the stability requirements, maintaining the original control of the operation of the air-conditioning unit.

[0013] According to the first aspect of the present application, whether the operation of the air-conditioning unit meets the stability requirement is determined in the following manner: whether the fluctuation amplitude of the acquired operation data of the air-conditioning unit within a predetermined time period is less than a fluctuation amplitude threshold.

[0014] According to the first aspect of the present application, in step S3, selectively performing auxiliary control on the operation of the air-conditioning unit includes: determining whether the acquired operation data of the air-conditioning unit reaches a predetermined similarity compared to the corresponding training data of the air-conditioning unit model; when the predetermined similarity is reached, performing auxiliary control on the operation of the air-conditioning unit to generate recommended input control data; when the predetermined similarity is not reached, maintaining the original control of the operation of the air-conditioning unit.

[0015] According to the first aspect of the present application, judging whether the acquired operating data of the air-conditioning unit reaches a predetermined similarity compared with the corresponding training data of the air-conditioning unit model includes: selecting several parameters from the input parameters and output parameters of the air-conditioning unit model, using the data corresponding to the selected several parameters as the training data of the similarity model, and training the similarity model based on the training data to obtain a determined similarity model; setting a predetermined similarity; selecting the operating data corresponding to the training data of the similarity model at each moment from the acquired operating data of the air-conditioning unit; inputting the selected operating data into the determined similarity model to obtain the calculated similarity; judging whether the obtained calculated similarity reaches the predetermined similarity.

[0016] According to the first aspect of the present application, the determined air-conditioning unit model is obtained in the following manner: determining the input parameters and output parameters of the air-conditioning unit model; obtaining the operating data of the air-conditioning unit corresponding to the input parameters and the output parameters; filtering the obtained operating data; generating the input data and output data of the air-conditioning unit model based on the filtered operating data; and using the generated input data and output data as training data of the air-conditioning unit model to train the air-conditioning unit model to obtain the determined air-conditioning unit model.

[0017] According to the first aspect of the present application, a queue container of a certain length is set for the training data of the air conditioning unit model, and the air conditioning unit model is trained based on the training data in the queue container. When new training data written into the air conditioning unit model causes the total training data to exceed the size of the queue container, the training data written earlier into the queue container is correspondingly overflowed from the queue container earlier.

[0018] According to a first aspect of the present application, an input convex neural network is used to obtain an air conditioning unit model.

[0019] According to the first aspect of the present application, auxiliary control can be selectively performed on the air-conditioning unit when the air-conditioning unit is started.

[0020] According to the first aspect of the present application, the air conditioning unit includes a compressor, an evaporator, a condenser and an economizer. The input parameters of the air conditioning unit model include: the compressor frequency of the air conditioning unit, the compressor intake guide vane opening, the condenser liquid level valve opening, the economizer liquid level valve opening, the evaporator inlet water temperature, the evaporator water flow, the condenser inlet water temperature and the condenser water flow. The output parameters of the air conditioning unit model include: the evaporator outlet water temperature, the condenser liquid level, the economizer liquid level, the evaporation pressure, the condensation pressure, the economizer pressure and the input power of the air conditioning unit.

[0021] According to the first aspect of the present application, in step S6, determining whether the updated operating data of the air-conditioning unit obtained meets the operating conditions includes: determining whether the updated input power of the air-conditioning unit is less than the input power before the operation of the air-conditioning unit is controlled based on the recommended input control data; determining whether surge occurs in the compressor of the air-conditioning unit; determining whether the updated evaporator pressure, condenser pressure, and economizer pressure of the air-conditioning unit exceed the maximum permissible pressure; determining whether there is ice formation based on the updated evaporating pressure of the evaporator of the air-conditioning unit; and determining whether the updated condenser liquid level and economizer liquid level of the air-conditioning unit are within the permissible liquid level range.

[0022] According to the first aspect of the present application, removing predicted output data and a portion of input data corresponding to the predicted output data according to safe operating rules includes removing the following data: the predicted evaporator outlet water temperature of an air-conditioning unit whose deviation from the evaporator outlet water temperature set value of the air-conditioning unit exceeds a deviation threshold; the predicted evaporator pressure, predicted condenser pressure and predicted economizer pressure of an air-conditioning unit that exceed a pressure threshold; the input compressor frequency of an air-conditioning unit that is less than the updated compressor surge frequency obtained for the air-conditioning unit; and the predicted condenser liquid level and predicted economizer liquid level of an air-conditioning unit that exceed a liquid level threshold.

[0023] According to the first aspect of the present application, in step S7, the original control of the operation of the air-conditioning unit is switched to and the operation of the air-conditioning unit is continuously monitored. When the operation of the air-conditioning unit is stable, the next control cycle is performed and the process goes to step S1.

[0024] According to a second aspect of the present application, the present application provides a controller for an air conditioning unit. The controller includes: a processor configured to execute the aforementioned control method to control the operation of the air conditioning unit.

[0025] According to a third aspect of the present application, the present application provides an air conditioning system, which includes an air conditioning unit, a detection device and a controller. The detection device is connected to the air conditioning unit and configured to obtain operating data of the air conditioning unit. The controller is connected to the air conditioning unit and the detection device and includes a processor. The processor is configured to execute the aforementioned control method based on the operating data of the air conditioning unit obtained from the detection device to control the operation of the air conditioning unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The drawings are not drawn to scale. In the drawings, each identical or nearly identical component represented in different figures is represented by the same reference numeral. For purposes of clarity, not every component may be labeled in every drawing. In the drawings:

[0027] Figure 1 shows a block diagram structure of an air conditioning system according to the present application;

[0028] Figure 2A Shows Figure 1 The overall flow chart of the control method of the air conditioning system shown;

[0029] Figure 2B Shows Figure 2A A flowchart of an embodiment of step 212 in the control method shown;

[0030] Figure 3A Shows Figure 2A A flowchart of a first embodiment of step 208 in the control method shown;

[0031] Figure 3B Shows Figure 2A A flowchart of a second embodiment of step 208 in the control method shown;

[0032] Figure 3C Shows Figure 3B A flowchart of an embodiment of step 312 is shown;

[0033] Figure 4A Shows Figure 2A A flowchart of an embodiment of step 210 in the control method shown;

[0034] Figure 4B Shows Figure 4A A flowchart of an embodiment of step 408 is shown;

[0035] Figure 5 Shows the acquisition Figure 2A A flowchart of an embodiment of a method for determining an air conditioning unit model in step 210 is shown;

[0036] Figure 6A schematic diagram showing the comparison of prediction results of the air conditioning unit model created by inputting a convex neural network (ICNN) and a deep neural network (DNN);

[0037] Figure 7 A block diagram of a queue container for accommodating training data of an air conditioning unit model is shown;

[0038] Figures 8A-8B Shows Shows Figure 2B A schematic diagram of an embodiment of the control execution result obtained in step 226 is shown; and

[0039] Fig. 9 Shows Figure 1 The structural block diagram of the controller is shown. DETAILED DESCRIPTION

[0040] Various specific embodiments of the present application will be described below with reference to the accompanying drawings which constitute a part of this specification. It should be understood that, where possible, the same or similar reference numerals used in this application refer to the same components.

[0041] Figure 1 FIG. 1 shows a block diagram of an air conditioning system 100 according to the present application. Figure 1 As shown, the air conditioning system 100 includes an air conditioning unit 101, a detection device 106 and a controller 110. The air conditioning unit 101 includes an evaporator 102, a compressor 103, a condenser 104 and an economizer 105. In other embodiments, the air conditioning unit 101 includes other suitable structures, such as an evaporator, a compressor and a condenser.

[0042] The air conditioning unit 101 is configured to operate to provide cooling or heat to the required space. The detection device 106 is connected to the air conditioning unit 101 and is configured to detect and obtain the operation data of the air conditioning unit 101 when it is in operation. In one embodiment, the detection device 106 includes a sensor built into the air conditioning unit 101 for obtaining the operation data of the corresponding components of the air conditioning unit 101. In other embodiments, the detection device 106 includes a sensor arranged outside the corresponding components of the air conditioning unit 101 for obtaining the operation data of the corresponding components. The controller 110 is connected to the detection device 106 by a connecting line 112, and obtains the operation data of each component of the air conditioning unit 101 detected by the detection device 106 by the connecting line 112. The controller 110 is configured to execute the control method of the present application based on the obtained operation parameters to output control data to the air conditioning unit 101, thereby controlling the operation of the air conditioning unit 101. The controller 110 is also connected to the air conditioning unit 101 through the connection line 111, and outputs control data to the air conditioning unit 101 through the connection line 111. In one embodiment, the control data can be output to the corresponding components of the air conditioning unit 101 to control the operation of these components. In other embodiments, the control data is output to other suitable components of the air conditioning unit 101 to control the operation of the air conditioning unit 101.

[0043] Figure 2A Shows Figure 1 The overall flow chart of the control method 200 of the air conditioning system 100 is shown.

[0044] like Figure 2A As shown, at step 202, the control method 200 of the air conditioning system 100 starts to be executed. Then, the process goes to step 204 from step 202. At step 204, the operation of the air conditioning unit 101 of the air conditioning system 100 is controlled. Then, the process goes to step 206 from step 204. At step 206, the operation data of the air conditioning unit 101 during operation is obtained in real time.

[0045] At step 208, based on the acquired operating data of the air-conditioning unit 101, the operation of the air-conditioning unit 101 is selectively auxiliary controlled. Selectively auxiliary controlling the operation of the air-conditioning unit 101 includes auxiliary controlling the operation of the air-conditioning unit 101 when the operation of the air-conditioning unit 101 satisfies a predetermined condition. In one embodiment, when the operation of the air-conditioning unit 101 is stable, auxiliary control of the operation of the air-conditioning unit 101 begins. In another embodiment, when the operating data of the air-conditioning unit 101 under the current working conditions is sufficiently similar to the corresponding training data of the air-conditioning unit model, auxiliary control of the operation of the air-conditioning unit 101 begins. See below for details on selectively auxiliary controlling the operation of the air-conditioning unit 101. Figure 3A-3B Then, the process goes to step 210 from step 208 .

[0046] At step 210, when auxiliary control is performed on the operation of the air-conditioning unit 101, recommended input control data is generated based on the determined air-conditioning unit model. The air-conditioning unit model is a model that predicts the performance of the air-conditioning unit. The air-conditioning unit model includes input parameters and output parameters, and the output parameters include, for example, input power, performance parameters (COP), etc. When the data corresponding to the input parameters are input into the determined air-conditioning unit model, the model can output (predict) the corresponding output parameter data, for example, predicting the performance of the air-conditioning unit under the working conditions corresponding to the input parameter data. In one embodiment, the recommended input control data is the input control data corresponding to the optimal predicted output data generated by the air-conditioning unit model based on the input control data, see for details. Figure 4A For example, when the predicted output data is the unit input power, controlling the operation of the air conditioning unit 101 based on the recommended input control data can enable the air conditioning unit to operate at the optimal unit input power, thereby achieving energy-saving optimization of the unit operation. Then, the process goes to step 212 from step 210.

[0047] At step 212, the operation of the air conditioning unit 101 is controlled based on the recommended input control data. Then, step 212 goes to step 214. At step 214, it is determined whether the air conditioning system 100 needs to be shut down. If it needs to be shut down, step 214 goes to step 216. Then, if it does not need to be shut down, the next control cycle is performed and step 214 goes to step 204 to control the air conditioning unit 101 to continue to operate under the next control cycle. At step 216, the execution of the control method 200 of the air conditioning system 100 is terminated. In one embodiment, the above-mentioned control method 200 of the air conditioning system 100 can be implemented after the air conditioning unit is started, for example, including selectively performing auxiliary control on the operation of the air conditioning unit 101.

[0048] Figure 2B Show Figure 2A FIG. 2 is a detailed flowchart of an embodiment of step 212 in the control method 200 shown in FIG. Figure 2A At step 212 shown, the operation of the air conditioning unit 101 is controlled based on the recommended input control data.

[0049] like Figure 2B As shown by Figure 2A Step 210 in Go to Figure 2BStep 220 in the process. At step 220, it is determined whether the recommended input control data is the same as the current operation data of the air-conditioning unit 101. As mentioned above, the recommended input control data is the input control data corresponding to the prediction output data generated by the air-conditioning unit model based on the input control data when the prediction output data is optimal. When the recommended input control data is the same as the current operation data of the air-conditioning unit 101, it indicates that the current operation data is already the optimal input control data, and the original control of the operation of the air-conditioning unit 101 can be maintained. When the recommended input control data is different from the current operation data of the air-conditioning unit 101, it indicates that the output data predicted by the air-conditioning unit model corresponding to the recommended input control data is more optimized than the current operation data, that is, more optimized input control data is found, and at this time, the original control of the operation of the air-conditioning unit 101 is switched to the control of the operation of the air-conditioning unit 101 based on the recommended updated input control data.

[0050] At step 220, when the recommended input control data is the same as the current operation data of the air conditioning unit 101, the process proceeds from step 220 to step 222. At step 222, the original control of the operation of the air conditioning unit 101 is maintained, and then the process proceeds from step 222 to step 223. Figure 2A . At step 204 in . At step 220, when the recommended input control data is different from the current operation data of the air-conditioning unit 101, the process proceeds from step 220 to step 224. At step 224, the operation of the air-conditioning unit 101 is controlled based on the recommended updated input control data. Then, the process proceeds from step 224 to step 226. In other embodiments, the original control of the operation of the air-conditioning unit 101 is maintained or the operation of the air-conditioning unit 101 is controlled based on the recommended updated input control data by other suitable means. At step 226, when the operation of the air-conditioning unit 101 is controlled based on the recommended updated input control data, the updated operation data of the air-conditioning unit 101 is obtained. Then, the process proceeds from step 226 to step 228.

[0051] At step 228, it is determined whether the updated operating data of the air conditioning unit 101 obtained at step 226 meets the operating conditions. When the updated operating data meets the operating conditions, the process goes to step 228. Figure 2A Step 214 in Figure 2A In step 214 of the embodiment, it is determined whether the air conditioning system 100 needs to be shut down. When the air conditioning system 100 does not need to be shut down, the next control cycle is performed and the control is turned to Figure 2AStep 204 in the above. When at least one of the updated operation data of the air-conditioning unit 101 obtained at step 226 does not meet the operation conditions, step 228 is transferred to step 230. At step 230, the auxiliary control of the operation of the air-conditioning unit 101 is switched to the original control of the operation of the air-conditioning unit 101. Then, step 230 is transferred to step 232. In one embodiment, for a two-stage centrifugal water-cooling unit, it is determined whether the updated input power is less than the input power before the recommended control parameter data is not executed; it is determined whether the compressor surges; it is determined according to the updated evaporator pressure, condenser pressure, and economizer pressure whether the structural safety meets the requirements after the recommended new control parameter data is executed, that is, whether the pressure exceeds the maximum allowable pressure; it is determined according to the updated evaporator evaporation pressure whether there will be icing; it is determined whether the updated condenser liquid level and economizer liquid level are within the allowable liquid level range. If all the above conditions are met, it indicates that the air-conditioning unit operates normally under the recommended input control data, and the next control cycle is repeated. If any of the above conditions is not met, it indicates that the air conditioning unit is not operating normally under the recommended input control data, and the original control of the air conditioning unit is restored. Figures 8A-8B The updated operation data generated by the air conditioning unit 101 running for a period of time based on the recommended updated input control data is shown, as described below. In other embodiments, it is determined whether other suitable operation parameter data meet the operation conditions to determine whether the air conditioning unit runs normally under the recommended input control data.

[0052] At step 232, when switching to the original control of the operation of the air conditioning unit 101, the operation of the air conditioning unit 101 is continued to be monitored, and it is determined whether the operation of the air conditioning unit 101 is stable. For example, the operation data of the air conditioning unit 101 is obtained, and based on the operation data, it is determined whether the air conditioning unit 101 is running stably. When the operation of the air conditioning unit 101 is unstable, the process goes to step 230 from step 232 to continue the original control of the operation of the air conditioning unit 101. When the operation of the air conditioning unit 101 is stable, the process goes to step 232 from step 230 to continue the original control of the operation of the air conditioning unit 101. Figure 2A Then, when the air conditioning system 100 does not need to be shut down, the next control cycle is performed and the control is turned to Figure 2AStep 204 in the process. That is, when the auxiliary control of the operation of the air-conditioning unit 101 is switched to the original control of the operation of the air-conditioning unit 101, the air-conditioning unit 101 is controlled to continue to operate under the next control cycle only after the original control of the air-conditioning unit 101 makes the operation of the air-conditioning unit 101 stable. Because after switching to the original control, the air-conditioning unit has fluctuations in control parameters and response parameters, immediately entering the next control cycle at this time may result in inaccurate operating data being obtained in the next control cycle, thereby, for example, causing inappropriate control of the operation of the air-conditioning unit. Therefore, the present application enters the next control cycle after the overall state parameters of the air-conditioning unit tend to be stable.

[0053] As described above, the present application provides an auxiliary control combined with the original control of the air-conditioning unit, which can realize seamless switching between the original control and the auxiliary control, that is, it does not affect the original control, and at the same time adds auxiliary control to make the operation of the air-conditioning unit more optimized, such as more energy-saving, and maintain performance over a long life cycle.

[0054] Figure 3A Show Figure 2A A detailed flowchart of a first embodiment of step 208 in the control method 200 shown in FIG. Figure 3B Show Figure 2A FIG. 2 is a detailed flowchart of a second embodiment of step 208 in the control method 200 shown in FIG. Figure 2A At step 208 shown, based on the acquired operation data of the air-conditioning unit 101, auxiliary control is selectively performed on the operation of the air-conditioning unit 101.

[0055] like Figure 3A As shown by Figure 2A Step 206 in Go to Figure 3A In step 302, based on the operation data of the air conditioning unit 101 obtained in step 206, it is determined whether the operation of the air conditioning unit 101 meets the stability requirement. When the operation of the air conditioning unit 101 does not meet the stability requirement, step 302 is transferred to step Figure 2A Step 204 in the above process is performed to maintain the original control of the operation of the air conditioning unit 101. When the operation of the air conditioning unit 101 meets the stability requirement, the process goes to step 304 from step 302. At step 304, the operation of the air conditioning unit 101 is auxiliary controlled. Then, the process goes to step 304. Figure 2AStep 210 in . In one embodiment, whether the operation of the air-conditioning unit 101 meets the stability requirement is determined based on the time series fluctuation amplitude of the operation data of the air-conditioning unit 101 and the fluctuation amplitude threshold, for example, based on whether the fluctuation amplitude of the acquired operation data of the air-conditioning unit 101 within a predetermined time period is less than the fluctuation amplitude threshold. In one embodiment, if the central value (median) of the operation data of the air-conditioning unit 101 within a 5-minute sliding window does not exceed 2% of the average value within the sliding window, it is determined that the operation of the air-conditioning unit 101 meets the stability requirement. In other embodiments, whether the operation of the air-conditioning unit 101 meets the stability requirement is determined by other suitable methods.

[0056] like Figure 3B As shown by Figure 2A Step 206 in Go to Figure 3B In step 312 of the embodiment, based on the operation data of the air conditioning unit 101 obtained in step 206, it is determined whether the operation data of the air conditioning unit 101 obtained in step 206 reaches a predetermined similarity with the corresponding training data of the air conditioning unit model. When the operation data of the air conditioning unit 101 obtained in step 206 does not reach a predetermined similarity with the corresponding training data of the air conditioning unit model, step 312 is transferred to step 313. Figure 2A Step 204 in the above example is performed to maintain the original control of the operation of the air conditioning unit 101. When the acquired operation data of the air conditioning unit 101 reaches a predetermined similarity with the corresponding training data of the air conditioning unit model, step 312 is transferred to step 314. At step 314, auxiliary control is performed on the operation of the air conditioning unit 101. Then, step 314 is transferred to step 315. Figure 2A Step 210 in .

[0057] The training data set of the air-conditioning unit model is difficult to include the operating data of the air-conditioning unit under special working conditions, because the prediction deviation of the air-conditioning unit model for these operating data is large, and it is not suitable to recommend new control parameter data. When the air-conditioning unit is operating under special working conditions, the present application maintains the original control of the air-conditioning unit to ensure the safe operation of the air-conditioning unit. In addition, when the air-conditioning unit is turned on and off, the setting parameters of the air-conditioning unit change greatly, and strong transient behavior will occur. At this time, the repeatability of the response is poor, and a huge amount of data is required for learning. When the operator makes a large change to the control parameters of the air-conditioning unit, similar problems mentioned above will occur. The present application starts to enter auxiliary control only when the operation of the air-conditioning unit meets the stability requirements and / or when the operating data of the acquired air-conditioning unit 101 reaches a predetermined similarity compared to the corresponding training data of the air-conditioning unit model, which can ensure the safe operation of the air-conditioning unit and reduce the large amount of training data required for the air-conditioning unit model. In other embodiments, the safe operation of the air-conditioning unit is ensured and the large amount of training data required for the air-conditioning unit model is reduced by other suitable methods.

[0058] Figure 3C Shows Figure 3B FIG. 3 is a flowchart of an embodiment of step 312 shown in FIG. Figure 3B At step 312 shown, it is determined whether the acquired operating data of the air-conditioning unit 101 reaches a predetermined similarity compared to the corresponding training data of the air-conditioning unit model.

[0059] like Figure 3C As shown by Figure 2A Step 206 in Go to Figure 3C In step 322, several parameters are selected from the input parameters and output parameters of the air conditioning unit model, and the operation data of the air conditioning unit 101 corresponding to the selected several parameters are used as training data of the similarity model, for example, the similarity model is a K nearest neighbor model (KNN), and the similarity model is trained based on these training data to obtain a determined similarity model. Then, the process goes to step 324 from step 322.

[0060] At step 324, a predetermined similarity is set. The similarity distance corresponds to the similarity. The smaller the similarity distance, the higher the similarity. In one embodiment, a predetermined similarity distance α is set. The predetermined similarity distance corresponds to the predetermined similarity. Then, step 324 is transferred to step 326. At step 326, operating data corresponding to the training data of the similarity model at each moment is selected from the acquired operating data of the air-conditioning unit 101. Then, step 326 is transferred to step 328. At step 328, the selected operating data is input into the determined similarity model to obtain the calculated similarity. In one embodiment, the calculated similarity distance is obtained, and the similarity distance is corresponded to the similarity. The similarity model is a model for calculating similarity. Then, step 328 is transferred to step 330.

[0061] At step 330, it is determined whether the calculated similarity obtained at step 328 reaches a predetermined similarity. When the calculated similarity obtained does not reach the predetermined similarity, the process proceeds from step 330 to step 328. Figure 2A Step 204 in the above embodiment is performed to maintain the original control of the operation of the air conditioning unit 101. When the calculated similarity obtained reaches the predetermined similarity, the process is transferred from step 330 to step 430. Figure 3B In step 314, auxiliary control is performed on the operation of the air conditioning unit. In other embodiments, the similarity between the acquired operation data of the air conditioning unit 101 and the corresponding training data of the air conditioning unit model is determined by other suitable methods.

[0062] In one embodiment, the evaporator inlet water temperature, evaporator outlet water temperature, evaporator water flow, condenser inlet water temperature, condenser outlet water temperature, and condenser water flow are selected from the training data set of the air conditioning unit model (including data of input parameters and output parameters) as training data, so as to create a similarity model, such as creating a K nearest neighbor model (KNN). A predetermined similarity distance α is set, for example, to 0.05. For the observed real-time operation data, the evaporator inlet water temperature, evaporator outlet water temperature, evaporator water flow, condenser inlet water temperature, condenser outlet water temperature, and condenser water flow corresponding to each time point are selected and substituted into the similarity model to calculate the similarity distance. If the calculated similarity distance is less than the predetermined similarity distance α, the air conditioning unit 101 is auxiliary controlled. Otherwise, the original control of the operation of the air conditioning unit 101 is maintained. In other embodiments, other suitable parameter data can be selected as training data for the similarity model.

[0063] Figure 4A Shows Figure 2A FIG. 2 is a flow chart of an embodiment of step 210 in the control method 200 shown in FIG. Figure 2A At step 210 shown, when performing auxiliary control on the operation of the air conditioning unit 101, recommended input control data is generated based on the determined air conditioning unit model.

[0064] like Figure 4A As shown by Figure 2A Step 208 in Go to Figure 4A . At step 402, one or more actual input data corresponding to at least a part of the input parameters of the air-conditioning unit model in the current operating data of the air-conditioning unit 101 are obtained. The actual input data obtained are data of adjustable control parameters of the unit. Adjusting these adjustable control parameters can adjust the operating conditions of the unit and adjust the actual output data or predicted data accordingly. Then, go to step 404 from step 402. At step 404, an optimized search range of the corresponding input parameters of the air-conditioning unit model is generated based on the one or more actual input data obtained. In one embodiment, an optimized search range of the corresponding input parameters of the air-conditioning unit model is generated in a neighboring interval of the one or more actual input data obtained. Then, go to step 406 from step 404.

[0065] At step 406, a plurality of input data in the optimization search range is obtained, and the obtained plurality of input data is input into the determined (trained) air conditioning unit model to generate a plurality of predicted output data. In one embodiment, the above-mentioned optimization search range is discretized to obtain a plurality of alternative control parameter data combinations. After all the alternative control parameter data combinations are normalized and input into the trained air conditioning unit model, the corresponding prediction results (i.e., predicted output data) are obtained. Then, step 406 is transferred to step 408.

[0066] At step 408, the optimized predicted output data and the optimized input data corresponding to the optimized predicted output data are obtained from the plurality of predicted output data, and the optimized input data is used as the recommended input control data. Then, step 408 is transferred to Figure 2A Step 212 in .

[0067] In one embodiment, an optimizer is run when auxiliary control is performed on the operation of the air-conditioning unit to give recommended control parameters based on the prediction results of the air-conditioning unit model. The present application proposes running a robust optimizer. Specifically, a search is made for more optimal control parameter data within a neighboring interval of the current control parameter data of the air-conditioning unit, where more optimal means, for example, minimizing the predicted input power while satisfying the predicted output water temperature and safety criteria. In other embodiments, more optimal includes optimizing other suitable predicted parameters while satisfying other suitable conditions. In one embodiment, a neighboring interval is set as follows:

[0068]

[0069] x Opening_IGV_j ∈[0.5x Opening_IGV0 , 1.5x Opening_IGV_0 ], i = 1, 2, ..., b

[0070] x Opening_EXV_Cond_m ∈[0.8x Opening_EXV_Cond_0 , 1.2x Opening_EXV_Cond_0 ], i = 1, 2, ..., c

[0071] x Opening_EXV_Eco_n ∈[0.8x Opening_EXV_Eco_0 , 1.2x Opening_EXVEco_0 ],i=1]2,…,d

[0072] where x Freq_Comp_i , x Opening_IGV_j , x Opening_EXV_Cond_m and x Opening_EXV_Eco_n represents the discrete values ​​of the control parameters compressor frequency, compressor inlet guide vane opening, condenser liquid level valve opening, and economizer liquid level valve opening within the above adjacent intervals. Freq_Comp_iThe discrete spacing d Freq_Comp =0.2Hz, x Opening_IGV_j The discrete spacing d Opening_IGV =2%, x Opening_EXV_Cond_m The discrete spacing d Opening_EXV_Cond =1%, x Opening_EXV_Eco_n The discrete spacing d Opening_EXV_Eco =1%. Therefore, the optimizer’s search space is x Freq_Comp_i , x Freq_Comp_i , x opening_EXV_Cond_m and x Opening_EXV_Eco_n In other embodiments, a more optimal control parameter is searched in other suitable adjacent intervals of other suitable control parameters.

[0073] The present application searches for better control parameter data in the vicinity of the current control parameter data, which can prevent the parameter data obtained through optimization from deviating too much from the current control parameter data, thereby causing operational safety problems. In addition, the search method can narrow the search range, increase the search density, improve the speed of extreme value search, and search for results that are closer to the extreme value.

[0074] Figure 4B Shows Figure 4A FIG. 4 is a flowchart of an embodiment of step 408 shown in FIG. Figure 4A At step 408 shown, the optimized predicted output data and the optimized input data corresponding to the optimized predicted output data are obtained from the plurality of predicted output data, and the optimized input data is used as the recommended input control data.

[0075] like Figure 4B As shown by Figure 4A Step 406 in Go to Figure 4B In step 412 of the embodiment, the plurality of predicted output data are compared with the corresponding actual output data to filter out the predicted output data that is less than the corresponding actual output data. In one embodiment, filtering out the predicted output data that is less than the corresponding actual output data includes filtering out the predicted input power that is less than the actual (measured) input power of the air conditioning unit 101. Then, the process proceeds from step 412 to step 414.

[0076] At step 414, a portion of the predicted output data and the input data corresponding to the predicted output data are removed according to the safe operation rule. Then, the process proceeds to step 416 from step 414. At step 416, the optimal predicted output data and the optimized input data corresponding to the optimal predicted output data are obtained from the predicted output data after screening and removal, and the optimized input data is used as the recommended control data. In one embodiment, obtaining the optimal predicted output data includes: obtaining the optimal predicted input power of the air conditioning unit 101.

[0077] In one embodiment, the deviation between the predicted data (e.g., input power) and the current measured data is calculated, all data whose predicted values ​​are less than the actual values ​​are screened out, and the corresponding control parameter data are arranged in rows in order of predicted values ​​from small to large. The following parameter data are removed according to the safety rules: the predicted evaporator outlet water temperature whose deviation from the evaporator outlet water temperature setting value exceeds the deviation threshold; the predicted evaporator pressure, predicted condenser pressure, and predicted economizer pressure that exceed the pressure threshold; the input compressor frequency that is less than the updated surge frequency obtained; and the predicted condenser liquid level and economizer liquid level that exceed the liquid level threshold. Based on the above screening and removal, several predicted output data are obtained, from which the optimal predicted output data is obtained, and the optimized input data corresponding to the optimal predicted output data is obtained as the recommended control data. In other embodiments, the recommended input control data is generated by other suitable methods.

[0078] exist Figure 2B In one embodiment, the Figure 4B Step 416 (which is a more specific step in step 210, i.e., obtaining recommended control data) is transferred to step 220. In one embodiment of step 220, after filtering and removing data at step 416, if the number of rows of the remaining data is less than 0, a null value is returned, indicating that no recommended control parameter data superior to the current control parameter data of the air-conditioning unit is obtained. At this time, the recommended input control data is the current control parameter data of the air-conditioning unit 101, so the original control of the operation of the air-conditioning unit 101 is maintained. If the number of rows of the remaining data is greater than 0, the first row of control parameter data with the smallest predicted output data (e.g., input power) is returned as the recommended control parameter data. At this time, the recommended control parameter data is different from the current control parameter data of the air-conditioning unit, and is an updated control parameter data, so the operation of the air-conditioning unit 101 is controlled based on the recommended updated input control data. In other embodiments, the original control of the operation of the air-conditioning unit 101 is maintained or the operation of the air-conditioning unit 101 is controlled based on the recommended updated input control data by other suitable means.

[0079] Figure 5 Shows the acquisition Figure 2A A flowchart of an embodiment of a method 500 for determining an air conditioning unit model in step 210 is shown.

[0080] like Figure 5 As shown, at step 502, the method 500 for obtaining the determined air conditioning unit model begins to execute. Then, step 502 goes to step 504. At step 504, the input parameters and output parameters of the air conditioning unit model are determined. Then, step 504 goes to step 506. At step 506, the operation data of the air conditioning unit 101 corresponding to the input parameters and output parameters determined at step 504 are obtained. Then, step 506 goes to step 508. At step 508, the operation data obtained at step 506 are filtered. Then, step 508 goes to step 510. At step 510, based on the filtered operation data at step 508, the input data and output data of the air conditioning unit model are generated. Then, step 510 goes to step 512. At step 512, the input data and output data generated at step 510 are used as training data of the air conditioning unit model to train the air conditioning unit model to obtain the determined air conditioning unit model. Then, step 512 goes to step 514. At step 514 , the execution of the method 500 for obtaining the determined air conditioning unit model ends.

[0081] The air conditioning unit model is a model for predicting the performance of the air conditioning unit. In one embodiment, the present application uses a neural network to create an air conditioning unit model. The neural network, as a universal function approximator, can be expressed as f(x, y; w, b), where x and y represent the input and output of the neural network, respectively, w represents the weight parameter, and b represents the bias parameter. In one embodiment, the present application uses an input convex neural network (ICNN) to create an air conditioning unit model. ICNN is a function approximator f ICNN (x, y; w, b), its characteristic is convex. The strict mathematical definition of a convex function is as follows:

[0082] If function f: R n →R is a convex function, then it satisfies the following two conditions:

[0083] 1) f is a convex set;

[0084] 2) For x∈X, y∈Y, 0≤α≤1, f(αx+(1-αy))≤αf(x)+(1-α)f(y).

[0085] Using an input convex neural network as an air conditioning unit model can significantly increase the interpretability of traditional black box models and the robustness of the control optimization process. The convexity of the convex function of ICNN makes mathematical optimization problems easier to solve. For example, in convex optimization, the local optimum must be the global optimum.

[0086] The specific structure and mathematical derivation of the input convex neural network can be found in the academic paper "Input Convex Neural Networks, Proceedings of the 34th International Conference on Machine Learning, Sydney, Australia, PMLR 70, 2017" by Brandon Amos et al. The paper provides a mathematical proof that if a neural network is convex under input x, it must be ensured that:

[0087] 1) The weight parameter w is non-negative;

[0088] 2) The activation function is convex and does not decrease as the input increases.

[0089] In one embodiment, the present application uses an input convex neural network to create an air conditioning unit model in the form of: i Non-negative, the inter-layer activation function is the linear rectification function (Rectified Linear Unit, ReLU), the mathematical expression is as follows:

[0090]

[0091] In one embodiment, the air conditioning unit 101 is a centrifugal chiller, such as a two-stage centrifugal water-cooling unit. In one embodiment, the input parameters of the air conditioning unit (two-stage centrifugal water-cooling unit) model include compressor frequency, compressor intake guide vane opening, condenser liquid level valve opening, economizer liquid level valve opening, evaporator inlet water temperature, evaporator water flow, condenser inlet water temperature and condenser water flow. The output parameters of the air conditioning unit (two-stage centrifugal water-cooling unit) model include evaporator outlet water temperature, condenser liquid level, economizer liquid level, evaporation pressure, condensing pressure, economizer pressure and input power.

[0092] Based on the above-determined input parameters and output parameters of the air-conditioning unit model, the present application also obtains actual operation data of the air-conditioning unit corresponding to these input parameters and output parameters of the air-conditioning unit model, such as actual operation input data and actual operation output data. Specifically, the historical operation data of the centrifugal chiller is obtained, and the operation data corresponding to the above-mentioned input parameters and output parameters of the air-conditioning unit model are selected therefrom. Considering the impact of seasonal periodic changes in outdoor ambient temperature on the cooling water circuit, that is, the seasonal fluctuations in the condenser inlet water temperature, the length of the historical data used to train the model needs to be one year or more. In one embodiment, if the centrifugal unit is a single-stage centrifugal compressor, that is, there is no economizer for replenishing air and increasing enthalpy, the above-determined parameters of the air-conditioning unit model do not include the economizer liquid level valve opening, the economizer liquid level and the economizer pressure.

[0093] Furthermore, the present application cleans the above selected historical operation data. In order to maximize the robustness of the control, the present application selects the results close to the steady state from the historical operation data, for example, by removing the following three types of data through data cleaning:

[0094] 1) Historical data of the shutdown phase. Some data acquisition systems in air-conditioning systems automatically collect data as long as the system is powered on, and do not have the function of autonomously identifying the operating status of the unit. This results in a large amount of useless shutdown data in the historical data, which needs to be removed.

[0095] 2) Data with large fluctuations during the startup and shutdown stages. During the startup or shutdown process of the unit, the control parameters of the system show a large climb or fall. The relevant state response parameters depend not only on the current control state, but also on the size of the initial value. Training an accurate dynamic model requires a huge amount of data, so it is extremely expensive to obtain enough training samples through the operation of the actual unit, so this application removes such data.

[0096] 3) Large fluctuations in normal operation. During normal operation, the unit still experiences large disturbances, which can cause significant fluctuations in related parameters. For example, a sudden drop in the set water outlet will cause the control system to automatically increase the compressor speed for loading. In the following period, the input power increases rapidly, while the heat exchange delay effect causes the evaporator water temperature to drop slowly. The corresponding instantaneous calculated coefficient of performance (COP) is much higher than the steady-state COP.

[0097] For the steady-state model, the above three types of data are noise data that seriously deviate from the steady-state results, so this application clears them through a cleaning algorithm.

[0098] After the above cleaning, the present application uses the cleaned operating data as the input parameter data and output parameter data of the created air-conditioning unit model as training data, and trains the air-conditioning unit model based on the training data to obtain a determined air-conditioning unit model.

[0099] Figure 6 A schematic diagram showing the comparison of prediction results of the air conditioning unit model created by inputting convex neural network (ICNN) and deep neural network (DNN).

[0100] like Figure 6 As shown in the figure, the comparison results between the actual test data, the prediction data based on the deep neural network (DNN), and the prediction data based on the input convex neural network (ICNN) are shown under the same network structure, the same training data, and the same number of training times, thereby showing the prediction advantage of the air conditioning unit model created by the input convex neural network (ICNN). Figure 6 As shown, during the dynamic load reduction process of the unit from 3:40 to 4:50, except that the prediction accuracy of ICNN for the economizer pressure P_Eco (kPa) is slightly lower than that of DNN, the rest are better than DNN in prediction stability and accuracy. In particular, the prediction results of the evaporator outlet water temperature (T_Out_Water_Evap (℃)) and input power (Power (kW)) directly determine the performance of the subsequent model-based predictive control algorithm. Although ICNN has a slight deviation in the prediction of the economizer pressure, it has met the control accuracy requirements. Because the purpose of predicting pressure parameters in this application is to ensure structural safety, that is, after executing a certain control parameter, it is necessary to predict whether the future response pressure exceeds the allowable limit pressure of the structural design. Compared with deep neural networks, input convex neural networks have better prediction smoothness, reliability and robustness.

[0101] Figure 7 The structural block diagram of the queue container for accommodating the training data of the air conditioning unit model is shown. The long-term operation of the unit will lead to wear of mechanical parts and overall performance degradation, which makes it impossible for the manufacturer to fully utilize the performance of the old unit based on the optimization control strategy (such as PID parameters) designed based on the new unit. Therefore, the present application proposes a model training data refresh strategy, so that the air conditioning unit model applied by the aforementioned auxiliary control method can be retrained according to the annual operation data of the air conditioning unit, so that the air conditioning unit model can better predict the current state of the unit.

[0102] like Figure 7As shown, the queue container for accommodating the training data of the air-conditioning unit model has a certain length, for example, it can accommodate approximately two years of training data. In one embodiment, data writing and model training are performed once a year. In other embodiments, data writing and model training are performed by other suitable means. In the present application, when new training data written to the air-conditioning unit model causes the total training data to exceed the size of the queue container, the training data written earlier in the queue container is correspondingly overflowed from the queue container earlier, thereby ensuring that the training data stored in the queue container represents the current unit performance as much as possible. For example Figure 7 As shown, when the training data of the n-1th year is written into the queue container, the previous training data of the n-3th year is overflowed from the queue container, and the training data of the n-2th year and the training data of the n-1th year are retained in the queue container. When the air-conditioning unit model is trained in the n-1th year, the air-conditioning unit model can be trained based on the training data of the n-2th year and the training data of the n-1th year in the queue container, or based on the training data of the n-1th year. When the training data of the nth year is written into the queue container, the previous training data of the n-2th year is overflowed from the queue container, and the training data of the n-1th year and the training data of the nth year are retained in the queue container. When the air-conditioning unit model is trained in the nth year, the air-conditioning unit model can be trained based on the training data of the n-1th year and the training data of the nth year in the queue container, or based on the training data of the nth year. In other embodiments, the queue container has other suitable lengths for accommodating other suitable amounts of training data.

[0103] Figures 8A-8B Shows Figure 2B FIG. 2 is a schematic diagram of an embodiment of the control execution result obtained in step 226. As mentioned above, Figure 2B At step 226 shown in FIG. 1 , the operating data of the air conditioning unit when it is operating under the recommended control data, that is, the execution result under the recommended control data, is obtained. Figures 8A-8B As shown in FIG. 1 , various operating parameters of a two-stage centrifugal water-cooling unit, namely, evaporator outlet water temperature, condenser and economizer liquid levels, anti-surge frequency, input power, COP, energy savings, compressor frequency, compressor inlet guide vane opening, condenser liquid level valve opening, economizer liquid level valve opening (from top to bottom), are plotted at a time period of 16200x10s=45h (hours). For example, it can be observed that the deviation between the predicted value of the evaporator outlet water temperature and the set value is less than 0.1°C, and the cumulative energy savings is >15%. These plots can be used to implement Figure 2B The operation at step 228 is to determine whether the updated operating data of the air-conditioning unit 101 obtained at step 226 meets the operating conditions.

[0104] Fig. 9 Shows Figure 1The structural block diagram of the controller 110 is shown in FIG. Fig. 9 As shown, the controller 110 includes a bus 901, a processor 902, a memory 903, an input interface 904, and an output interface 905. The processor 902, the memory 903, the input interface 904, and the output interface 905 are connected to the bus 901. The processor 902 can read a program (or instruction) from the memory 903, and execute the program (or instruction) to perform the processing of data. The processor 902 can also write data or a program (or instruction) into the memory 903. The memory 903 can store programs (instructions) or data. By executing the instructions in the memory 903, the processor 902 can control the memory 903, the input interface 904, and the output interface 905.

[0105] The input interface 904 is configured to receive the operating parameter data of the air conditioning unit 101 acquired by the detection device 106 through the connection line 112. The input interface 904 is also configured to convert the received parameter data into data recognizable by the processor 902 and output the data to the processor 902.

[0106] The processor 902 is configured to process (e.g., calculate) the received data to generate a control signal. In one embodiment, the processor 902 processes the acquired operating parameter data of the air conditioning unit 101 to perform the steps in the aforementioned method, such as generating a signal to maintain the original control of the unit, or obtaining recommended updated control data and generating a control signal for the air conditioning unit to operate under the recommended updated control data.

[0107] The output interface 905 is configured to receive a control signal from the processor 902, convert the control signal into a signal of the air conditioning unit 101, and send the control signal to the air conditioning unit 101 to control the operation of the air conditioning unit 101. In one embodiment, the output interface 905 sends the control signal to the air conditioning unit 101 through the connection line 111 to control the operation of the air conditioning unit 101.

[0108] All features disclosed in this specification (including any attached claims, abstracts and drawings) and / or all steps of any method or process disclosed in this specification can be combined in any suitable combination, except for at least some mutually exclusive combinations of such features and / or steps. The application is not limited to the specific order of steps described in the specification, but includes other suitable order of steps for implementing the application.

[0109] Although the present application has been described in conjunction with the examples of the embodiments summarized above, it is obvious to those of ordinary skill in the art that various alternatives, modifications, changes, improvements and / or substantially equivalent solutions, whether known or currently or foreseeable in the near future, may be apparent. In addition, the technical effects and / or technical problems described in this specification are exemplary rather than restrictive; so the disclosure in this specification may be used to solve other technical problems and have other technical effects and / or may solve other technical problems. Therefore, the examples of the embodiments of the present application as stated above are intended to be illustrative rather than restrictive. Various changes may be made without departing from the spirit or scope of the present application. Therefore, the present application is intended to include all known or earlier developed alternatives, modifications, changes, improvements and / or substantially equivalent solutions.

Claims

1. A control method (200) for an air conditioning unit (101), the control method (200) comprising: S1: controlling the operation of the air conditioning unit (101); S2: acquiring the operating data of the air conditioning unit (101) and the air conditioning unit model in real time; S3: Based on the acquired operation data of the air-conditioning unit (101), selectively performing auxiliary control on the operation of the air-conditioning unit (101); S4: when performing auxiliary control on the operation of the air conditioning unit (101), generating recommended input control data based on the determined model of the air conditioning unit; S5: Controlling the operation of the air conditioning unit (101) based on the recommended input control data.

2. The control method (200) for the air conditioning unit (101) according to claim 1, wherein: In step S5: When the recommended input control data is the same as the current operation data of the air conditioning unit (101), the original control of the operation of the air conditioning unit (101) is maintained and the process goes to step S1; When the recommended input control data is different from the current operation data of the air conditioning unit (101), the operation of the air conditioning unit (101) is controlled based on the recommended updated input control data.

3. The control method (200) for the air conditioning unit (101) according to claim 2, the control method (200) further comprising: S6: when controlling the operation of the air conditioning unit (101) based on the recommended updated input control data, obtaining updated operation data of the air conditioning unit (101) to determine whether the updated operation data satisfies the operation conditions; S7: If the updated operating data all meet the operating conditions, the next control cycle is performed and the process goes to step S1; otherwise, the process switches to the original control of the operation of the air-conditioning unit (101).

4. The control method (200) for the air conditioning unit (101) according to claim 1, wherein: In step S4, the recommended input control data is generated by running the optimizer, wherein running the optimizer includes: S4.1: Obtaining one or more actual input data corresponding to at least a part of input parameters of the air conditioning unit model from the current operation data of the air conditioning unit (101); S4.2: generating an optimization search range of a corresponding input parameter of the air conditioning unit model based on the acquired one or more actual input data; S4.3: Acquire a plurality of input data in the optimization search range, and input the acquired plurality of input data into the air conditioning unit model to generate a plurality of predicted output data; S4.4: Obtain optimized predicted output data and optimized input data corresponding to the optimized predicted output data from the plurality of predicted output data, and use the optimized input data as the recommended input control data.

5. The control method (200) for the air conditioning unit (101) according to claim 4, wherein: In the step S4.4, obtaining the optimized predicted output data and the optimized input data corresponding to the optimized predicted output data includes: Comparing the plurality of predicted output data with corresponding actual output data to filter out the predicted output data that is smaller than the corresponding actual output data; removing the predicted output data and a portion of the input data corresponding to the predicted output data according to a safe operation rule; The optimal predicted output data and the optimized input data corresponding to the optimal predicted output data are obtained from the predicted output data after the screening and the removal.

6. The control method (200) for the air conditioning unit (101) according to claim 5, wherein: In step S4.4, Screening out the predicted output data that is smaller than the corresponding actual output data comprises: screening out the predicted input power that is smaller than the actual input power of the air conditioning unit (101); Acquiring the optimal predicted output data includes: acquiring the optimal predicted input power of the air conditioning unit (101).

7. The control method (200) for the air conditioning unit (101) according to claim 1, wherein: In the step S3, selectively performing auxiliary control on the operation of the air conditioning unit (101) includes: Based on the acquired operating data of the air-conditioning unit (101), determining whether the operation of the air-conditioning unit (101) meets the stability requirement; When the operation of the air conditioning unit (101) meets the stability requirement, auxiliary control is performed on the operation of the air conditioning unit (101) to generate recommended input control data; When the operation of the air-conditioning unit (101) does not meet the stability requirement, the original control of the operation of the air-conditioning unit (101) is maintained.

8. The control method (200) for the air conditioning unit (101) according to claim 7, wherein: Whether the operation of the air-conditioning unit (101) meets the stability requirement is judged in the following manner: whether the fluctuation amplitude of the acquired operation data of the air-conditioning unit (101) within a predetermined time period is less than a fluctuation amplitude threshold.

9. The control method (200) for the air conditioning unit (101) according to claim 1, wherein: In the step S3, selectively performing auxiliary control on the operation of the air conditioning unit (101) includes: Determining whether the acquired operating data of the air-conditioning unit (101) reaches a predetermined similarity with the corresponding training data of the air-conditioning unit model; When a predetermined similarity is reached, auxiliary control is performed on the operation of the air conditioning unit (101) to generate recommended input control data; When the predetermined similarity is not reached, the original control of the operation of the air-conditioning unit (101) is maintained.

10. The control method (200) for the air conditioning unit (101) according to claim 9, wherein: Determining whether the acquired operating data of the air-conditioning unit (101) reaches a predetermined similarity compared to the corresponding training data of the air-conditioning unit model comprises: Selecting several parameters from the input parameters and output parameters of the air conditioning unit model, using data corresponding to the selected several parameters as training data of the similarity model, and training the similarity model based on the training data to obtain a determined similarity model; Set a predetermined similarity; Selecting, from the acquired operating data of the air conditioning unit (101), operating data corresponding to the training data of the similarity model at each moment; inputting the selected operating data into the determined similarity model to obtain a calculated similarity; It is determined whether the obtained calculated similarity reaches the predetermined similarity.

11. The control method (200) for the air conditioning unit (101) according to claim 1, wherein: The determined air conditioning unit model is obtained in the following manner: Determining input parameters and output parameters of the air conditioning unit model; Acquiring operation data of the air conditioning unit (101) corresponding to the input parameter and the output parameter; Filtering the acquired operation data; generating input data and output data of the air conditioning unit model based on the filtered operating data; The generated input data and the output data are used as training data of the air conditioning unit model to train the air conditioning unit model, so as to obtain the determined air conditioning unit model.

12. The control method (200) for the air conditioning unit (101) according to claim 11, wherein: A queue container of a certain length is set for the training data of the air conditioning unit model, and the air conditioning unit model is trained based on the training data in the queue container. When new training data written into the air conditioning unit model causes the total training data to exceed the size of the queue container, the training data written earlier into the queue container is correspondingly overflowed from the queue container earlier.

13. The control method (200) for the air conditioning unit (101) according to claim 1, wherein: An input convex neural network is used to obtain the air conditioning unit model.

14. The control method (200) for the air conditioning unit (101) according to claim 1, wherein: When the air conditioning unit (101) is started, auxiliary control can be selectively performed on the air conditioning unit (101).

15. The control method (200) for an air conditioning unit (101) according to claim 1, wherein the air conditioning unit (101) comprises a compressor (103), an evaporator (102), a condenser (104) and an economizer (105), wherein: The input parameters of the air conditioning unit model include: the compressor frequency of the air conditioning unit (101), the compressor intake guide vane opening, the condenser liquid level valve opening, the economizer liquid level valve opening, the evaporator inlet water temperature, the evaporator water flow, the condenser inlet water temperature and the condenser water flow, The output parameters of the air conditioning unit model include: the evaporator water outlet temperature, condenser liquid level, economizer liquid level, evaporation pressure, condensation pressure, economizer pressure and input power of the air conditioning unit (101).

16. The control method (200) for the air conditioning unit (101) according to claim 3, wherein: In step S6, judging whether the updated operating data of the air-conditioning unit (101) obtained satisfies the operating conditions includes: determining whether the updated input power of the air conditioning unit (101) is less than the input power before the operation of the air conditioning unit (101) is controlled based on the recommended input control data; Determining whether a compressor of the air conditioning unit (101) is experiencing surge; Determining whether the updated evaporator pressure, condenser pressure, and economizer pressure of the air conditioning unit (101) exceed the maximum permissible pressure; determining whether there is icing according to the updated evaporation pressure of the evaporator of the air conditioning unit (101); and It is determined whether the updated condenser liquid level and economizer liquid level of the air conditioning unit (101) are within a permissible liquid level range.

17. The control method (200) for the air conditioning unit (101) according to claim 5, wherein: Removing the predicted output data and a portion of the input data corresponding to the predicted output data according to the safe operation rule includes removing the following data: a predicted evaporator outlet water temperature of the air conditioning unit (101) whose deviation from a set value of the evaporator outlet water temperature of the air conditioning unit (101) exceeds a deviation threshold; a predicted evaporator pressure, a predicted condenser pressure, and a predicted economizer pressure of the air conditioning unit (101) that exceed a pressure threshold; an input compressor frequency of the air-conditioning unit (101) that is less than the acquired updated compressor surge frequency of the air-conditioning unit (101); as well as A predicted condenser liquid level and a predicted economizer liquid level of the air conditioning unit (101) that exceed liquid level thresholds.

18. The control method (200) for the air conditioning unit (101) according to claim 3, wherein: In step S7, the original control of the operation of the air-conditioning unit (101) is switched to and the operation of the air-conditioning unit (101) is continuously monitored. When the operation of the air-conditioning unit (101) is stable, the next control cycle is performed and the process goes to step S1.

19. A controller (110) for an air conditioning unit, the controller (110) comprising: A processor (902), wherein the processor (902) is configured to execute the control method described in claims 1-18 to control the operation of the air conditioning unit (101).

20. An air conditioning system (100), the air conditioning system (100) comprising: Air conditioning unit (101); a detection device (106), the detection device (106) being connected to the air conditioning unit (101) and configured to obtain operating data of the air conditioning unit (101); and A controller (110), the controller (110) being connected to the air conditioning unit (101) and the detection device (106), the controller (110) comprising a processor (902), Wherein, the processor (902) is configured to execute the control method described in claims 1 to 18 based on the operating data of the air-conditioning unit (101) obtained from the detection device (106) to control the operation of the air-conditioning unit (101).

Images