Fault diagnosis method based on artificial intelligence and ultra-low temperature freezer
Through the fault diagnosis method based on artificial intelligence, the temperature and power prediction model is used to solve the problem that ultra-low temperature refrigerators are difficult to diagnose faults, and the accurate judgment of refrigeration effect and compressor control and the determination of the cause of faults are achieved, which improves the accuracy of fault diagnosis. Sex and efficiency.
Patent Information
- Application Number
- CN202411063851.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-05
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-08-05
AI Technical Summary
It is difficult to detect faults and determine the cause of the fault in ultra-low temperature refrigerators, and the existing technology cannot effectively diagnose its refrigeration effect and compressor control problems.
Using artificial intelligence-based fault diagnosis method, by obtaining the temperature data in the intermediate heat exchanger and refrigerator and the power data of the compressor, using the temperature prediction model and power prediction model, calculate the temperature and power matching scores, judge the fault and determine the main reason.
It improves the accuracy of fault diagnosis, can judge whether the ultra-low temperature refrigerator has a fault from the two aspects of refrigeration effect and compressor control, and determines the cause of the fault, improving the accuracy and efficiency of temperature and power prediction.
Smart Images

Figure CN118583324B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of refrigeration technology, and in particular to an artificial intelligence-based fault diagnosis method and an ultra-low temperature freezer. Background Art
[0002] In the related art, although the refrigeration capacity of ultra-low temperature freezers is stronger than that of ordinary freezers, their structure is also more complicated, and it is more difficult to detect whether they are faulty and to determine the cause of the failure.
[0003] The information disclosed in the background technology section of this application is only intended to deepen the understanding of the general background technology of this application, and should not be regarded as an admission or any form of suggestion that the information constitutes the prior art already known to those skilled in the art. Summary of the invention
[0004] The present invention provides a fault diagnosis method and an ultra-low temperature freezer based on artificial intelligence, which can solve the technical problem that it is difficult to find out whether the ultra-low temperature freezer has a fault and it is also difficult to determine the cause of the fault in the related technology.
[0005] According to a first aspect of the present invention, there is provided a fault diagnosis method based on artificial intelligence, comprising:
[0006] At multiple moments in the i-th monitoring cycle, first temperature data in the intermediate heat exchanger and second temperature data in the cabinet of the ultra-low temperature freezer are respectively obtained, where i is a positive integer;
[0007] Obtaining a first refrigeration power of the primary compressor and a second refrigeration power of the secondary compressor in the i-th monitoring cycle, wherein the primary compressor is used in a primary refrigeration cycle, the primary refrigeration cycle is used to transfer heat from the intermediate heat exchanger to the air outside the ultra-low temperature freezer, and the secondary compressor is used in a secondary refrigeration cycle, the secondary refrigeration cycle is used to transfer heat in the space inside the cabinet of the ultra-low temperature freezer to the intermediate heat exchanger;
[0008] The first refrigeration power of the first-stage compressor in the i-th monitoring cycle, the second refrigeration power of the second-stage compressor, the first temperature data and the second temperature data at the first moment in the i-th monitoring cycle are input into the trained temperature prediction model to obtain the first predicted temperature data in the intermediate heat exchanger at multiple moments in the i-th monitoring cycle, and the second predicted temperature data in the cabinet of the ultra-low temperature freezer;
[0009] Obtaining a temperature matching score according to the first predicted temperature data, the second predicted temperature data, the first temperature data, and the second temperature data;
[0010] Input the first temperature data and the second temperature data at the first moment in the i-th monitoring cycle into the trained power prediction model to obtain the first predicted refrigeration power of the first-stage compressor and the second predicted refrigeration power of the second-stage compressor in the i-th monitoring cycle;
[0011] Input the first predicted refrigeration power, the second predicted refrigeration power, the first temperature data at the first moment in the i-th monitoring cycle, and the second temperature data into the trained temperature prediction model to obtain the first verification temperature data in the intermediate heat exchanger at multiple moments in the i-th monitoring cycle, and the second verification temperature data in the cabinet of the ultra-low temperature freezer;
[0012] When the first verification temperature data and the second verification temperature data meet preset requirements, determining a power matching score according to the first predicted cooling power, the second predicted cooling power, the first cooling power, the second cooling power and the temperature matching score;
[0013] A fault score of the ultra-low temperature freezer is obtained according to the power matching score and the temperature matching score.
[0014] According to a second aspect of the present invention, there is provided an ultra-low temperature freezer, comprising: a primary compressor, a secondary compressor, an intermediate heat exchanger and a controller, wherein the primary compressor is used in a primary refrigeration cycle, the primary refrigeration cycle is used to transfer heat from the intermediate heat exchanger to air outside the ultra-low temperature freezer, and the secondary compressor is used in a secondary refrigeration cycle, the secondary refrigeration cycle is used to transfer heat in the space inside the cabinet of the ultra-low temperature freezer to the intermediate heat exchanger;
[0015] The controller is used to:
[0016] At multiple moments in the i-th monitoring cycle, first temperature data in the intermediate heat exchanger and second temperature data in the cabinet of the ultra-low temperature freezer are respectively obtained, where i is a positive integer;
[0017] Obtaining a first refrigeration power of the first-stage compressor and a second refrigeration power of the second-stage compressor in the i-th monitoring cycle;
[0018] The first refrigeration power of the first-stage compressor in the i-th monitoring cycle, the second refrigeration power of the second-stage compressor, the first temperature data and the second temperature data at the first moment in the i-th monitoring cycle are input into the trained temperature prediction model to obtain the first predicted temperature data in the intermediate heat exchanger at multiple moments in the i-th monitoring cycle, and the second predicted temperature data in the cabinet of the ultra-low temperature freezer;
[0019] Obtaining a temperature matching score according to the first predicted temperature data, the second predicted temperature data, the first temperature data, and the second temperature data;
[0020] Input the first temperature data and the second temperature data at the first moment in the i-th monitoring cycle into the trained power prediction model to obtain the first predicted refrigeration power of the first-stage compressor and the second predicted refrigeration power of the second-stage compressor in the i-th monitoring cycle;
[0021] Input the first predicted refrigeration power, the second predicted refrigeration power, the first temperature data at the first moment in the i-th monitoring cycle, and the second temperature data into the trained temperature prediction model to obtain the first verification temperature data in the intermediate heat exchanger at multiple moments in the i-th monitoring cycle, and the second verification temperature data in the cabinet of the ultra-low temperature freezer;
[0022] When the first verification temperature data and the second verification temperature data meet preset requirements, determining a power matching score according to the first predicted cooling power, the second predicted cooling power, the first cooling power, the second cooling power and the temperature matching score;
[0023] A fault score of the ultra-low temperature freezer is obtained according to the power matching score and the temperature matching score.
[0024] Technical effect: According to the present invention, the deviation between the measured temperature data and the expected temperature of the ultra-low temperature freezer can be judged based on the temperature prediction model and the temperature data and power measured during the monitoring period, so as to judge whether there is a problem with the refrigeration effect. On the other hand, the deviation between the actual power and the expected power of the ultra-low temperature freezer can be judged through the power prediction model, so as to judge whether there is a problem with the setting of the refrigeration power of the ultra-low temperature freezer. Then, it can be judged whether the ultra-low temperature freezer has a fault from the two aspects of the refrigeration effect and the control of the compressor, and the main cause of the fault can be determined to improve the accuracy of fault diagnosis. When training the temperature prediction model, the temperature stability of the intermediate heat exchanger, the relative difference between the first sample temperature data and the second sample temperature data, the average change rate of the second sample temperature data and other factors can be considered in the loss function, so as to improve the prediction accuracy of the temperature data of the monitoring period with greater monitoring significance in a targeted manner, improve the accuracy of the temperature prediction model, and improve the training efficiency. When determining the temperature matching score, the temperature matching score can be solved by the matching degree between the measured first temperature data and the expected value of the temperature data in the intermediate heat exchanger, and the matching degree between the second temperature data and the expected value of the temperature data in the cabinet, so that the temperature matching score can accurately describe the matching degree between the actual refrigeration effect of the ultra-low temperature freezer and the expected refrigeration effect, thereby improving the accuracy and objectivity of the temperature matching score. When training the power prediction model, the refrigeration effect that can be achieved by the predicted power output by the power prediction model can be considered in the loss function, so as to use the trained temperature prediction model to assist in training, thereby specifically improving the predicted data with poor refrigeration effect, improving the accuracy of the power prediction model, and improving the training efficiency. When determining the power matching score, the power matching score can be obtained by the matching degree between the actual power and the expected power of the ultra-low temperature freezer, and the matching degree between the refrigeration effect achieved by the actual power and the expected refrigeration effect, thereby improving the accuracy and comprehensiveness of the power matching score. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other embodiments can be obtained based on these drawings without creative work.
[0026] Figure 1 A schematic diagram of a flow chart of a fault diagnosis method based on artificial intelligence according to an embodiment of the present invention is exemplarily shown;
[0027] Figure 2 A schematic diagram of an ultra-low temperature freezer according to an embodiment of the present invention is exemplarily shown. DETAILED DESCRIPTION
[0028] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0029] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.
[0030] Figure 1 A flowchart of a fault diagnosis method based on artificial intelligence according to an embodiment of the present invention is exemplarily shown, and the method includes:
[0031] Step S101, at multiple moments in the i-th monitoring cycle, respectively obtain first temperature data in the intermediate heat exchanger and second temperature data in the cabinet of the ultra-low temperature freezer, where i is a positive integer greater than 1;
[0032] Step S102, obtaining a first refrigeration power of the primary compressor and a second refrigeration power of the secondary compressor in the i-th monitoring cycle, wherein the primary compressor is used in a primary refrigeration cycle, the primary refrigeration cycle is used to transfer heat from the intermediate heat exchanger to the air outside the ultra-low temperature freezer, and the secondary compressor is used in a secondary refrigeration cycle, the secondary refrigeration cycle is used to transfer heat in the space inside the cabinet of the ultra-low temperature freezer to the intermediate heat exchanger;
[0033] Step S103, inputting the first refrigeration power of the first-stage compressor in the i-th monitoring cycle, the second refrigeration power of the second-stage compressor, the first temperature data and the second temperature data at the first moment in the i-th monitoring cycle into the trained temperature prediction model, and obtaining the first predicted temperature data in the intermediate heat exchanger at multiple moments in the i-th monitoring cycle, and the second predicted temperature data in the cabinet of the ultra-low temperature freezer;
[0034] Step S104, obtaining a temperature matching score according to the first predicted temperature data, the second predicted temperature data, the first temperature data, and the second temperature data;
[0035] Step S105, inputting the first temperature data and the second temperature data at the first moment in the i-th monitoring cycle into the trained power prediction model to obtain the first predicted refrigeration power of the first-stage compressor and the second predicted refrigeration power of the second-stage compressor in the i-th monitoring cycle;
[0036] Step S106, inputting the first predicted refrigeration power, the second predicted refrigeration power, the first temperature data and the second temperature data at the first moment in the i-th monitoring cycle into the trained temperature prediction model, and obtaining the first verification temperature data in the intermediate heat exchanger at multiple moments in the i-th monitoring cycle, and the second verification temperature data in the cabinet of the ultra-low temperature freezer;
[0037] Step S107, when the first verification temperature data and the second verification temperature data meet the preset requirements, determining the power matching score according to the first predicted cooling power, the second predicted cooling power, the first cooling power, the second cooling power and the temperature matching score;
[0038] Step S108, obtaining a fault score of the ultra-low temperature freezer according to the power matching score and the temperature matching score.
[0039] According to the artificial intelligence-based fault diagnosis method of an embodiment of the present invention, the deviation between the measured temperature data of the ultra-low temperature freezer and the expected temperature can be judged based on the temperature prediction model and the temperature data and power measured during the monitoring period, thereby judging whether there is a problem with the refrigeration effect. On the other hand, the deviation between the actual power and the expected power of the ultra-low temperature freezer can also be judged through the power prediction model, thereby judging whether there is a problem with the refrigeration power setting of the ultra-low temperature freezer. Furthermore, it can be judged whether the ultra-low temperature freezer has a fault from the two aspects of the refrigeration effect and the control of the compressor, and the main cause of the fault can be determined, thereby improving the accuracy of fault diagnosis.
[0040] According to one embodiment of the present invention, in step S101, after the ultra-low temperature freezer is operated for a period of time, the first temperature data in the intermediate heat exchanger of the ultra-low temperature freezer and the second temperature data in the cabinet of the ultra-low temperature freezer can be detected. For example, after the ultra-low temperature freezer is operated stably, the above temperature data can be detected at multiple times in the i-th monitoring cycle.
[0041] According to one embodiment of the present invention, in step S102, the first refrigeration power and the second refrigeration power in each monitoring cycle remain unchanged, and the first refrigeration power of the first-stage compressor and the second refrigeration power of the second-stage compressor in the i-th monitoring cycle can be obtained, so that the expected temperature data in the ultra-low temperature freezer under the action of the first refrigeration power and the second refrigeration power can be determined, and then it can be determined whether the measured temperature data is consistent with the expected temperature data.
[0042] According to one embodiment of the present invention, in step S103, the first predicted temperature data expected in the intermediate heat exchanger and the second predicted temperature data expected in the cabinet of the ultra-low temperature freezer under the action of the first refrigeration power of the first compressor and the second refrigeration power of the second compressor can be predicted by the trained temperature prediction model. For example, the first refrigeration power, the second refrigeration power, the first temperature data and the second temperature data at the first moment in the i-th monitoring cycle can be input into the trained temperature prediction model, and the trained temperature prediction model can calculate how the temperature in the intermediate heat exchanger will change relative to the first temperature data at the first moment under the action of the first refrigeration power and the second refrigeration power, and how the temperature in the cabinet of the ultra-low temperature freezer will change relative to the second temperature data at the first moment, thereby obtaining the first predicted temperature data in the intermediate heat exchanger at multiple moments in the i-th monitoring cycle and the second predicted temperature data in the cabinet of the ultra-low temperature freezer.
[0043] According to one embodiment of the present invention, the first refrigeration power and the second refrigeration power will affect the temperature change rate in the intermediate heat exchanger and the temperature change rate in the cabinet of the ultra-low temperature freezer in a monitoring cycle. In addition, the first temperature data and the second temperature data at the first moment will also affect the above-mentioned temperature change rate. For example, the secondary refrigeration cycle conducts the heat in the cabinet of the ultra-low temperature freezer to the intermediate heat exchanger, and the first temperature data is higher than the second temperature data. The greater the difference between the first temperature data and the second temperature data, the more difficult it is to conduct heat, and the lower the temperature change rate. In addition, the second temperature data itself will also affect the above-mentioned temperature change rate. For example, the lower the second temperature data, the more difficult it is for the temperature in the cabinet of the ultra-low temperature freezer to continue to drop, and the lower the temperature change rate.
[0044] According to one embodiment of the present invention, therefore, based on the influence of the above multiple factors, the first predicted temperature data in the intermediate heat exchanger at multiple moments in the i-th monitoring cycle and the second predicted temperature data in the cabinet of the ultra-low temperature freezer can be predicted based on the temperature prediction model. The temperature prediction model is a deep learning neural network model, which can be calculated based on multiple factors to obtain accurate solution results. Before using the temperature prediction model to perform the above prediction operation, the temperature prediction model can be trained.
[0045] According to one embodiment of the present invention, the training step of the temperature prediction model includes: selecting sample ultra-low temperature freezers of the same model as the ultra-low temperature freezer; obtaining first sample temperature data of the intermediate heat exchanger and second sample temperature data inside the cabinet of multiple sample ultra-low temperature freezers at multiple moments in multiple monitoring cycles; obtaining first sample refrigeration power of the first-stage compressor and second sample refrigeration power of the second-stage compressor of multiple sample ultra-low temperature freezers in multiple monitoring cycles; inputting the first sample refrigeration power of the first-stage compressor and the second sample refrigeration power of the second-stage compressor of the sample ultra-low temperature freezer, as well as the first sample temperature data and the second sample temperature data at the first moment of the monitoring cycle, into the temperature prediction model to obtain the first predicted sample temperature data of the intermediate heat exchanger and the second predicted sample temperature data inside the cabinet at multiple moments in the monitoring cycle; obtaining the loss function of the temperature prediction model based on the first sample temperature data, the first predicted sample temperature data, the second sample temperature data and the second predicted sample temperature data; training the temperature prediction model based on the loss function of the temperature prediction model to obtain a trained temperature prediction model.
[0046] According to one embodiment of the present invention, in order to obtain more training data so that the temperature prediction model can undergo more training of the first sample cooling power and the second sample cooling power, as well as the first sample temperature data and the second sample temperature data, so that the temperature prediction model can make accurate predictions in a variety of situations, multiple sample ultra-low temperature freezers of the same model as the ultra-low temperature freezer can be selected, and the first sample temperature data of the intermediate heat exchanger and the second sample temperature data in the cabinet of the multiple sample ultra-low temperature freezers, as well as the first sample cooling power and the second sample cooling power in multiple monitoring cycles are actually measured at multiple times in multiple monitoring cycles, so as to obtain more training data.
[0047] According to one embodiment of the present invention, the first sample refrigeration power and the second sample refrigeration power of the sample ultra-low temperature freezer within a monitoring period, as well as the first sample temperature data and the second sample temperature data at the first moment of a monitoring period, can be input into a temperature prediction model. The temperature prediction model can calculate how the temperature in the intermediate heat exchanger changes relative to the first sample temperature data at the first moment under the action of the first sample refrigeration power and the second sample refrigeration power, and how the temperature in the cabinet changes relative to the second sample temperature data at the first moment, that is, the first predicted sample temperature data of the intermediate heat exchanger and the second predicted sample temperature data in the cabinet at multiple moments in the monitoring period are obtained.
[0048] According to one embodiment of the present invention, the loss function of the temperature prediction model can be obtained based on the error between the predicted temperature (i.e., the first predicted sample temperature data) and the actual measured temperature (i.e., the first sample temperature data) of the intermediate heat exchanger during the monitoring period, and the error between the predicted temperature (i.e., the second predicted sample temperature data) and the actual measured temperature (i.e., the second sample temperature data) in the cabinet, thereby training the temperature prediction model.
[0049] According to an embodiment of the present invention, according to the first sample temperature data, the first predicted sample temperature data, the second sample temperature data and the second predicted sample temperature data, a loss function of the temperature prediction model is obtained, including: according to formula (1), obtaining the loss function of the temperature prediction model ,
[0050] (1)
[0051] in, is the jth sample ultra-low temperature freezer in the kth monitoring cycle The first sample temperature data at a moment, is the jth sample ultra-low temperature freezer in the kth monitoring cycle The first predicted sample temperature data at the moment, n is the number of moments in the monitoring cycle, is the jth sample ultra-low temperature freezer in the kth monitoring cycle The first sample temperature data at a moment, is the time interval between adjacent moments, m is the number of monitoring cycles, N is the number of sample ultra-low temperature freezers, The jth sample ultra-low temperature freezer is the first in the kth monitoring cycle The second sample temperature data at the moment, is the jth sample ultra-low temperature freezer in the kth monitoring cycle The second predicted sample temperature data at the moment, The jth sample ultra-low temperature freezer is the first in the kth monitoring cycle The second sample temperature data at the moment, is the preset maximum temperature change rate, and is the preset weight, k≤m, j≤N, ,and , n, k, m, j and N are all positive integers.
[0052] According to one embodiment of the present invention, in formula (1), is the relative difference between the first sample temperature data and the first predicted sample temperature data, It is the average value of the above relative difference within a monitoring period. It is the average change rate of the first sample temperature data within a monitoring cycle. The larger the average change rate, the more unstable the temperature in the intermediate heat exchanger is within the monitoring cycle. The temperature in the intermediate heat exchanger should theoretically be stably maintained at a low level so that it can stably receive the heat conducted in the cabinet. If the temperature in the intermediate heat exchanger is unstable, the heat conduction efficiency from the cabinet to the intermediate heat exchanger is unstable, and it is difficult to effectively cool the cabinet. Therefore, if the average change rate is larger, the monitoring significance of the first sample temperature data is greater, that is, it is more necessary to focus and accurately monitor. Therefore, the above average change rate can be used as the weight of the average value of the relative difference to obtain the average value of the weighted relative difference of each monitoring cycle. .
[0053] According to one embodiment of the present invention, in each monitoring cycle, the sample ultra-low temperature freezer operates more stably in the later monitoring cycle, and the reference significance of the data collected in the monitoring cycle is greater. Therefore, the sequence number of the detection cycle can be used to weight the average value of the weighted relative difference, that is, using As the weight of the average value of the weighted relative difference of each monitoring cycle, the average value of the weighted relative difference of each sample ultra-low temperature freezer in multiple monitoring cycles is solved. Since the model of each sample ultra-low temperature freezer is the same, the average value of the weighted relative difference of each sample ultra-low temperature freezer in multiple monitoring cycles can be directly averaged to obtain the first term of the loss function of the temperature prediction model. During the training process, this item can be minimized, and based on the above weights, the prediction accuracy of monitoring cycles with unstable operation and greater monitoring significance can be improved in a targeted manner, thereby improving the prediction accuracy of the temperature prediction model for the first prediction sample temperature data.
[0054] According to one embodiment of the present invention, similar to the above, It is the average value of the relative difference between the second predicted sample temperature data and the second sample temperature data within a monitoring period. is the average change rate of the second sample temperature data within a monitoring cycle. Since the lower the second sample temperature data, the more difficult it is to conduct the heat in the cabinet to the intermediate heat exchanger, the smaller the change rate of the second sample temperature data. Therefore, the smaller the change rate of the second sample temperature data, the closer the second sample temperature data is to the set temperature of the cabinet, and the greater the monitoring significance of the second sample temperature data, that is, the more focused and accurate the monitoring is. Therefore, As the weight of the average value of the relative difference, the preset maximum temperature change rate may be the average change rate of the second sample temperature data in the first monitoring cycle, and the present invention does not limit the specific value of the preset maximum temperature change rate. is the relative difference between the first sample temperature data and the second sample temperature data at the first moment in the monitoring cycle. The larger the relative difference, the more difficult it is to transfer the heat in the cabinet to the intermediate heat exchanger, and the greater the monitoring significance of the second sample temperature data, that is, the more focused and accurate the monitoring is. Therefore, the relative difference can also be used as the weight of the average value of the relative difference. As the weight of the average of the relative differences, The meaning of is the same as above and will not be repeated here. After obtaining the above weights, the average value of the above relative difference of each monitoring cycle of each sample ultra-low temperature freezer can be processed by weighted average to obtain the second term of the loss function of the temperature prediction model: During the training process, this item can be minimized, and based on the above weights, the prediction accuracy of the monitoring period that is close to the set temperature in the cabinet and has greater monitoring significance can be improved in a targeted manner, thereby improving the prediction accuracy of the temperature prediction model for the second prediction sample temperature data.
[0055] According to one embodiment of the present invention, the above two items can be weightedly summed to obtain the loss function of the temperature prediction model, and then the loss function of the temperature prediction model is used for back propagation to adjust the parameters of the temperature prediction model to reduce the loss function. After multiple trainings, a trained temperature prediction model can be obtained.
[0056] In this way, the temperature stability of the intermediate heat exchanger, the relative difference between the first sample temperature data and the second sample temperature data, the average change rate of the second sample temperature data and other factors can be considered in the loss function, so as to specifically improve the prediction accuracy of the temperature data of the monitoring period with greater monitoring significance, improve the accuracy of the temperature prediction model, and improve the training efficiency.
[0057] According to an embodiment of the present invention, in step S104, obtaining a temperature matching score according to the first predicted temperature data, the second predicted temperature data, the first temperature data and the second temperature data includes: obtaining a temperature matching score according to formula (2): ,
[0058] (2)
[0059] in, is the first temperature data at the hth moment in the i-th monitoring cycle, is the first predicted temperature data at the hth moment in the i-th monitoring cycle, is the second temperature data at the hth moment in the i-th monitoring cycle, is the second predicted temperature data at the hth moment in the ith monitoring cycle, m is the number of monitoring cycles, h≤m, and both h and m are positive integers.
[0060] According to one embodiment of the present invention, in formula (2), is the relative difference between the measured first temperature data and the first predicted temperature data output by the trained temperature prediction model (which can be used as the expected value of the temperature data in the intermediate heat exchanger), is the average value of the relative difference within a monitoring period, and 1 minus the average value is the matching degree between the measured first temperature data and the expected value of the temperature data in the intermediate heat exchanger. Similarly, That is, the matching degree between the second temperature data and the expected value of the temperature data in the cabinet. The above two matching degrees are multiplied to obtain the temperature matching score, which can describe the overall matching degree between the measured temperature in the intermediate heat exchanger and the cabinet of the ultra-low temperature freezer and the expected value. If the temperature matching score is low, the gap between the measured temperature and the expected value is large, and there is a fault in the refrigeration cycle or the insulation of the freezer, for example, lack of refrigerant, or poor insulation performance, etc., resulting in difficulty in achieving the refrigeration effect corresponding to a certain refrigeration power under the action of the refrigeration power.
[0061] In this way, the temperature matching score can be solved by the matching degree between the measured first temperature data and the expected value of the temperature data in the intermediate heat exchanger, and the matching degree between the second temperature data and the expected value of the temperature data in the cabinet. The temperature matching score can accurately describe the matching degree between the actual refrigeration effect of the ultra-low temperature freezer and the expected refrigeration effect, thereby improving the accuracy and objectivity of the temperature matching score.
[0062] According to one embodiment of the present invention, in step S105, the power prediction model may be a deep learning neural network model, which may provide a suitable first predicted cooling power and a second predicted cooling power based on the first temperature data and the second temperature data at the first moment in the monitoring cycle, wherein the first predicted cooling power can keep the temperature in the intermediate heat exchanger stably at a low level, thereby improving the efficiency of the process of transferring heat from the cabinet to the intermediate heat exchanger, and the second predicted cooling power can maintain a good cooling effect in the cabinet, for example, maintaining a certain temperature drop rate. As described above, the cooling effect is affected by many factors, and therefore, a plurality of training data may be used to train the power prediction model, so that the power prediction model can provide accurate first predicted cooling power and second predicted cooling power based on many factors.
[0063] According to one embodiment of the present invention, the training step of the power prediction model includes: selecting sample ultra-low temperature freezers of the same model as the ultra-low temperature freezer; obtaining the third sample temperature data of the intermediate heat exchanger of the multiple sample ultra-low temperature freezers and the fourth sample temperature data in the cabinet at multiple moments in multiple monitoring cycles; obtaining the third sample refrigeration power of the first-stage compressor and the fourth sample refrigeration power of the second-stage compressor of the multiple sample ultra-low temperature freezers in multiple monitoring cycles; inputting the third sample temperature data and the fourth sample temperature data of the multiple sample ultra-low temperature freezers at the first moment in multiple monitoring cycles into the power prediction model to obtain the first sample predicted power of the first-stage compressor and the second sample predicted power of the second-stage compressor; and inputting the first sample predicted power, the fourth sample predicted power and the third sample predicted power into the power prediction model. The second sample predicted power, and the third sample temperature data and the fourth sample temperature data at the first moment in the monitoring period are input into the trained temperature prediction model to obtain the third predicted sample temperature data in the intermediate heat exchanger and the fourth predicted sample temperature data in the cabinet at multiple moments in the monitoring period; according to the first sample predicted power, the second sample predicted power, the third sample refrigeration power, the fourth sample refrigeration power, the third sample temperature data, the fourth sample temperature data, the third predicted sample temperature data and the fourth predicted sample temperature data, the loss function of the power prediction model is obtained; according to the loss function of the power prediction model, the power prediction model is trained to obtain the trained power prediction model.
[0064] According to one embodiment of the present invention, in order to obtain more training data, multiple sample ultra-low temperature freezers of the same model as the ultra-low temperature freezer can be used to collect training data. That is, the third sample temperature data of the intermediate heat exchanger and the fourth sample temperature data in the cabinet of multiple sample ultra-low temperature freezers, as well as the third sample refrigeration power of the first compressor and the fourth sample refrigeration power of the second compressor are obtained. The third sample temperature data and the fourth sample temperature data at the first moment in the monitoring cycle can be input into the power prediction model, and the first sample predicted power of the first compressor and the second sample predicted power of the second compressor are calculated by the power prediction model. The first sample predicted power and the second sample predicted power may not be appropriate powers. Therefore, the first sample predicted power, the second sample predicted power, and the third sample temperature data and the fourth sample temperature data at the first moment in the monitoring cycle can be input into the trained temperature prediction model. The trained temperature prediction model determines what kind of cooling effect can be achieved under the action of the first sample predicted power and the second sample predicted power, that is, how the temperature in the intermediate heat exchanger and the cabinet is expected to change relative to the third sample temperature data and the fourth sample temperature data at the first moment, and obtains the third predicted sample temperature data in the intermediate heat exchanger and the fourth predicted sample temperature data in the cabinet. After obtaining the above data, the loss function of the power prediction model can be calculated.
[0065] According to an embodiment of the present invention, according to the first sample predicted power, the second sample predicted power, the third sample cooling power, the fourth sample cooling power, the third sample temperature data, the fourth sample temperature data, the third predicted sample temperature data and the fourth predicted sample temperature data, obtaining the loss function of the power prediction model includes: according to formula (3), obtaining the loss function of the power prediction model ,
[0066] (3)
[0067] in, The predicted power of the first sample of the j-th sample ultra-low temperature freezer in the k-th monitoring period, is the third sample refrigeration power of the jth sample ultra-low temperature freezer in the kth monitoring cycle, is the jth sample ultra-low temperature freezer in the kth monitoring cycle The third predicted sample temperature data at the moment, The jth sample ultra-low temperature freezer is the first in the kth monitoring cycle The third sample temperature data at the moment, The predicted power of the second sample of the j-th sample ultra-low temperature freezer in the k-th monitoring period, is the fourth sample refrigeration power of the jth sample ultra-low temperature freezer in the kth monitoring cycle, is the jth sample ultra-low temperature freezer in the kth monitoring cycle The fourth predicted sample temperature data at the moment, is the jth sample ultra-low temperature freezer in the kth monitoring cycle The cooling power of the fourth sample at the moment, n is the number of moments in the monitoring cycle, m is the number of monitoring cycles, N is the number of sample ultra-low temperature freezers, and is the preset weight, k≤m, j≤N, ,and , n, k, m, j and N are all positive integers.
[0068] According to one embodiment of the present invention, is the relative deviation between the predicted power of the first sample and the cooling power of the third sample, It represents the average relative deviation between the third predicted sample temperature data and the third sample temperature data. The larger the average relative deviation is, the more difficult it is for the first sample predicted power output by the power prediction model to achieve the expected cooling effect. Therefore, it is necessary to increase the training intensity. As The weight of The higher the weight. The meaning of is the same as above and will not be repeated here. Further, the above weights can be used to perform a weighted average of the relative deviations between the first sample predicted power and the third sample refrigeration power of each sample ultra-low temperature freezer in each monitoring period to obtain the first term of the loss function of the power prediction model: , so that during the training process, the prediction data with poor refrigeration effect can be trained in a targeted manner to quickly improve the model accuracy. Similarly, the second term of the loss function of the power prediction model is the weighted average of the relative deviations between the second sample prediction power and the fourth sample refrigeration power of each sample ultra-low temperature freezer in each monitoring period, which will not be repeated here.
[0069] According to one embodiment of the present invention, the above two items can be weightedly summed to obtain the loss function of the power prediction model, and then the loss function of the power prediction model can be used for back propagation to adjust the parameters of the power prediction model to reduce the loss function. After multiple trainings, a trained power prediction model can be obtained.
[0070] In this way, the cooling effect that can be achieved by the predicted power output by the power prediction model can be considered in the loss function, so that the trained temperature prediction model can be used to assist training, thereby specifically improving the predicted data with poor cooling effect, improving the accuracy of the power prediction model, and improving training efficiency.
[0071] According to one embodiment of the present invention, in step S106, the first predicted refrigeration power and the second predicted refrigeration power output by the trained power prediction model are powers that can keep the ultra-low temperature freezer have a good refrigeration effect. In this step, the refrigeration effects of the first predicted refrigeration power and the second predicted refrigeration power can be verified by the trained temperature prediction model. The first predicted refrigeration power, the second predicted refrigeration power, the first temperature data and the second temperature data at the first moment in the i-th monitoring cycle are input into the trained temperature prediction model to obtain the first verification temperature data in the intermediate heat exchanger at multiple moments in the i-th monitoring cycle and the second verification temperature data in the cabinet of the ultra-low temperature freezer, and to determine whether the first verification temperature data can remain stable relative to the first temperature data at the first moment, and whether the second verification temperature data can maintain a sufficient rate of change relative to the second temperature data at the first moment.
[0072] According to one embodiment of the present invention, in step S107, if the first verification temperature data and the second verification temperature data meet the preset requirements, that is, whether the first verification temperature data can remain stable relative to the first temperature data at the first moment, and whether the second verification temperature data can maintain a sufficient rate of change relative to the second temperature data at the first moment, then the power matching score can be determined based on the first predicted cooling power, the second predicted cooling power, the first cooling power, the second cooling power and the temperature matching score.
[0073] According to an embodiment of the present invention, when the first verification temperature data and the second verification temperature data meet the preset requirements, determining the power matching score according to the first predicted cooling power, the second predicted cooling power, the first cooling power and the second cooling power includes: determining the power matching score according to formula (4): ,
[0074] (4)
[0075] in, Score the temperature match, is the first cooling power, is the first predicted cooling power, is the second cooling power, The second predicted cooling power.
[0076] According to one embodiment of the present invention, in formula (4), is the relative deviation between the first refrigeration power and the first predicted refrigeration power, is the matching degree between the first cooling power and the first predicted cooling power. Similarly, It is the matching degree between the second refrigeration power and the second predicted refrigeration power. The two are multiplied and can be used to describe the overall matching degree between the first refrigeration power and the second refrigeration power of the ultra-low temperature freezer and the expected refrigeration power. In addition, the temperature matching score can be used to describe the matching degree between the refrigeration effect of the ultra-low temperature freezer and the expected refrigeration effect. If the matching degree is less than 1, it is due to the deviation between the first refrigeration power of the ultra-low temperature freezer and the first predicted refrigeration power, or the deviation between the second refrigeration power and the second predicted refrigeration power. Therefore, on the basis of the multiplication of the above two, it can be multiplied by the temperature matching score to obtain the power matching score, which is used to describe the matching degree between the actual power of the ultra-low temperature freezer and the expected power, as well as the matching degree between the refrigeration effect achieved by the actual power and the expected refrigeration effect. If the power matching score is, it means that there is a problem with the power setting of the ultra-low temperature freezer.
[0077] In this way, the power matching score can be obtained through the matching degree between the actual power and the expected power of the ultra-low temperature freezer, and the matching degree between the cooling effect achieved by the actual power and the expected cooling effect, thereby improving the accuracy and comprehensiveness of the power matching score.
[0078] According to one embodiment of the present invention, in step S108, the power matching score and the temperature matching score may be weightedly summed to obtain a fault score of the ultra-low temperature freezer. If the fault score of the ultra-low temperature freezer is low, it indicates that there is a fault with the ultra-low temperature freezer, which may include lack of refrigerant, poor thermal insulation performance of the cabinet, problems with temperature setting and control, etc., which may prompt the staff to perform repairs.
[0079] According to the fault diagnosis method based on artificial intelligence of the embodiment of the present invention, the deviation between the measured temperature data of the ultra-low temperature freezer and the expected temperature can be judged based on the temperature prediction model and the temperature data and power measured during the monitoring period, so as to judge whether there is a problem with the refrigeration effect. On the other hand, the deviation between the actual power of the ultra-low temperature freezer and the expected power can be judged through the power prediction model, so as to judge whether there is a problem with the setting of the refrigeration power of the ultra-low temperature freezer. Then, it can be judged whether the ultra-low temperature freezer has a fault from the two aspects of the refrigeration effect and the control of the compressor, and the main cause of the fault can be determined, so as to improve the accuracy of fault diagnosis. When training the temperature prediction model, the temperature stability of the intermediate heat exchanger, the relative difference between the first sample temperature data and the second sample temperature data, the average change rate of the second sample temperature data and other factors can be considered in the loss function, so as to improve the prediction accuracy of the temperature data of the monitoring period with greater monitoring significance in a targeted manner, improve the accuracy of the temperature prediction model, and improve the training efficiency. When determining the temperature matching score, the temperature matching score can be solved by the matching degree between the measured first temperature data and the expected value of the temperature data in the intermediate heat exchanger, and the matching degree between the second temperature data and the expected value of the temperature data in the cabinet, so that the temperature matching score can accurately describe the matching degree between the actual refrigeration effect of the ultra-low temperature freezer and the expected refrigeration effect, thereby improving the accuracy and objectivity of the temperature matching score. When training the power prediction model, the refrigeration effect that can be achieved by the predicted power output by the power prediction model can be considered in the loss function, so as to use the trained temperature prediction model to assist in training, thereby specifically improving the predicted data with poor refrigeration effect, improving the accuracy of the power prediction model, and improving the training efficiency. When determining the power matching score, the power matching score can be obtained by the matching degree between the actual power and the expected power of the ultra-low temperature freezer, and the matching degree between the refrigeration effect achieved by the actual power and the expected refrigeration effect, thereby improving the accuracy and comprehensiveness of the power matching score.
[0080] Figure 2 A schematic diagram of an ultra-low temperature freezer according to an embodiment of the present invention is exemplarily shown, wherein the ultra-low temperature freezer comprises: a primary compressor, a secondary compressor, an intermediate heat exchanger and a controller, wherein the primary compressor is used in a primary refrigeration cycle, the primary refrigeration cycle is used to transfer heat from the intermediate heat exchanger to the air outside the ultra-low temperature freezer, and the secondary compressor is used in a secondary refrigeration cycle, the secondary refrigeration cycle is used to transfer heat in the space inside the cabinet of the ultra-low temperature freezer to the intermediate heat exchanger;
[0081] The controller is used to:
[0082] At multiple moments in the i-th monitoring cycle, first temperature data in the intermediate heat exchanger and second temperature data in the cabinet of the ultra-low temperature freezer are respectively obtained, where i is a positive integer;
[0083] Obtaining a first refrigeration power of the first-stage compressor and a second refrigeration power of the second-stage compressor in the i-th monitoring cycle;
[0084] The first refrigeration power of the first-stage compressor in the i-th monitoring cycle, the second refrigeration power of the second-stage compressor, the first temperature data and the second temperature data at the first moment in the i-th monitoring cycle are input into the trained temperature prediction model to obtain the first predicted temperature data in the intermediate heat exchanger at multiple moments in the i-th monitoring cycle, and the second predicted temperature data in the cabinet of the ultra-low temperature freezer;
[0085] Obtaining a temperature matching score according to the first predicted temperature data, the second predicted temperature data, the first temperature data, and the second temperature data;
[0086] Input the first temperature data and the second temperature data at the first moment in the i-th monitoring cycle into the trained power prediction model to obtain the first predicted refrigeration power of the first-stage compressor and the second predicted refrigeration power of the second-stage compressor in the i-th monitoring cycle;
[0087] Input the first predicted refrigeration power, the second predicted refrigeration power, the first temperature data at the first moment in the i-th monitoring cycle, and the second temperature data into the trained temperature prediction model to obtain the first verification temperature data in the intermediate heat exchanger at multiple moments in the i-th monitoring cycle, and the second verification temperature data in the cabinet of the ultra-low temperature freezer;
[0088] When the first verification temperature data and the second verification temperature data meet preset requirements, determining a power matching score according to the first predicted cooling power, the second predicted cooling power, the first cooling power, the second cooling power and the temperature matching score;
[0089] A fault score of the ultra-low temperature freezer is obtained according to the power matching score and the temperature matching score.
[0090] It should be understood by those skilled in the art that the embodiments of the present invention described above and shown in the accompanying drawings are only examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functional and structural principles of the present invention have been demonstrated and explained in the embodiments, and the embodiments of the present invention may be deformed or modified in any way without departing from the principles.
[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A fault diagnosis method based on artificial intelligence, characterized in that: include: At multiple moments in the i-th monitoring cycle, first temperature data in the intermediate heat exchanger and second temperature data in the cabinet of the ultra-low temperature freezer are respectively obtained, where i is a positive integer; Obtaining a first refrigeration power of the primary compressor and a second refrigeration power of the secondary compressor in the i-th monitoring cycle, wherein the primary compressor is used in a primary refrigeration cycle, the primary refrigeration cycle is used to transfer heat from the intermediate heat exchanger to the air outside the ultra-low temperature freezer, and the secondary compressor is used in a secondary refrigeration cycle, the secondary refrigeration cycle is used to transfer heat in the space inside the cabinet of the ultra-low temperature freezer to the intermediate heat exchanger; The first refrigeration power of the first-stage compressor in the i-th monitoring cycle, the second refrigeration power of the second-stage compressor, the first temperature data and the second temperature data at the first moment in the i-th monitoring cycle are input into the trained temperature prediction model to obtain the first predicted temperature data in the intermediate heat exchanger at multiple moments in the i-th monitoring cycle, and the second predicted temperature data in the cabinet of the ultra-low temperature freezer; Obtaining a temperature matching score according to the first predicted temperature data, the second predicted temperature data, the first temperature data, and the second temperature data; Input the first temperature data and the second temperature data at the first moment in the i-th monitoring cycle into the trained power prediction model to obtain the first predicted refrigeration power of the first-stage compressor and the second predicted refrigeration power of the second-stage compressor in the i-th monitoring cycle; Input the first predicted refrigeration power, the second predicted refrigeration power, the first temperature data at the first moment in the i-th monitoring cycle, and the second temperature data into the trained temperature prediction model to obtain the first verification temperature data in the intermediate heat exchanger at multiple moments in the i-th monitoring cycle, and the second verification temperature data in the cabinet of the ultra-low temperature freezer; In the case where the first verification temperature data and the second verification temperature data meet the preset requirements, the power matching score is determined according to the first predicted cooling power, the second predicted cooling power, the first cooling power, the second cooling power and the temperature matching score; wherein the preset requirements include that the first verification temperature data can remain stable relative to the first temperature data at the first moment, and that the second verification temperature data can maintain a sufficient rate of change relative to the second temperature data at the first moment; A fault score of the ultra-low temperature freezer is obtained according to the power matching score and the temperature matching score.
2. The fault diagnosis method based on artificial intelligence according to claim 1 is characterized in that: The training steps of the temperature prediction model include: Select a sample ultra-low temperature freezer of the same model as the ultra-low temperature freezer; At multiple moments in multiple monitoring cycles, obtaining first sample temperature data of the intermediate heat exchangers of multiple sample ultra-low temperature freezers and second sample temperature data in the cabinets; Obtaining first sample refrigeration powers of the primary compressor and second sample refrigeration powers of the secondary compressor of multiple sample ultra-low temperature freezers in multiple monitoring cycles; Input the first sample refrigeration power of the first compressor and the second sample refrigeration power of the second compressor of the sample ultra-low temperature freezer, as well as the first sample temperature data and the second sample temperature data at the first moment of the monitoring period, into the temperature prediction model to obtain the first predicted sample temperature data of the intermediate heat exchanger and the second predicted sample temperature data in the cabinet at multiple moments in the monitoring period; Obtaining a loss function of a temperature prediction model according to the first sample temperature data, the first predicted sample temperature data, the second sample temperature data, and the second predicted sample temperature data; The temperature prediction model is trained according to the loss function of the temperature prediction model to obtain a trained temperature prediction model.
3. The fault diagnosis method based on artificial intelligence according to claim 2, characterized in that: Obtaining a loss function of a temperature prediction model according to the first sample temperature data, the first predicted sample temperature data, the second sample temperature data, and the second predicted sample temperature data, including: According to the formula , Obtain the loss function of the temperature prediction model ,in, is the jth sample ultra-low temperature freezer in the kth monitoring cycle The first sample temperature data at a moment, is the jth sample ultra-low temperature freezer in the kth monitoring cycle The first predicted sample temperature data at the moment, n is the number of moments in the monitoring cycle, is the jth sample ultra-low temperature freezer in the kth monitoring cycle The first sample temperature data at a moment, is the time interval between adjacent moments, m is the number of monitoring cycles, N is the number of sample ultra-low temperature freezers, The jth sample ultra-low temperature freezer is the first in the kth monitoring cycle The second sample temperature data at the moment, is the jth sample ultra-low temperature freezer in the kth monitoring cycle The second predicted sample temperature data at the moment, The jth sample ultra-low temperature freezer is the first in the kth monitoring cycle The second sample temperature data at the moment, is the preset maximum temperature change rate, and is the preset weight, k≤m, j≤N, ,and , n, k, m, j and N are all positive integers.
4. The fault diagnosis method based on artificial intelligence according to claim 1 is characterized in that: The training steps of the power prediction model include: Select a sample ultra-low temperature freezer of the same model as the ultra-low temperature freezer; At multiple moments in multiple monitoring cycles, obtaining third sample temperature data of the intermediate heat exchanger of multiple sample ultra-low temperature freezers and fourth sample temperature data in the cabinet; Obtaining a third sample refrigeration power of a primary compressor and a fourth sample refrigeration power of a secondary compressor of a plurality of sample ultra-low temperature freezers in a plurality of monitoring cycles; Input the third sample temperature data and the fourth sample temperature data of the plurality of sample ultra-low temperature freezers at the first moment in the plurality of monitoring cycles into the power prediction model to obtain the first sample predicted power of the first-stage compressor and the second sample predicted power of the second-stage compressor; Input the first sample predicted power, the second sample predicted power, and the third sample temperature data and the fourth sample temperature data at the first moment in the monitoring period into the trained temperature prediction model, and obtain the third predicted sample temperature data in the intermediate heat exchanger and the fourth predicted sample temperature data in the cabinet at multiple moments in the monitoring period; Obtaining a loss function of a power prediction model according to the first sample predicted power, the second sample predicted power, the third sample cooling power, the fourth sample cooling power, the third sample temperature data, the fourth sample temperature data, the third predicted sample temperature data, and the fourth predicted sample temperature data; The power prediction model is trained according to the loss function of the power prediction model to obtain a trained power prediction model.
5. The fault diagnosis method based on artificial intelligence according to claim 4, characterized in that: Obtaining a loss function of a power prediction model according to the first sample predicted power, the second sample predicted power, the third sample cooling power, the fourth sample cooling power, the third sample temperature data, the fourth sample temperature data, the third predicted sample temperature data, and the fourth predicted sample temperature data, including: According to the formula , Obtain the loss function of the power prediction model ,in, The predicted power of the first sample of the j-th sample ultra-low temperature freezer in the k-th monitoring period, is the third sample refrigeration power of the jth sample ultra-low temperature freezer in the kth monitoring cycle, is the jth sample ultra-low temperature freezer in the kth monitoring cycle The third predicted sample temperature data at the moment, The jth sample ultra-low temperature freezer is the first in the kth monitoring cycle The third sample temperature data at the moment, The predicted power of the second sample of the j-th sample ultra-low temperature freezer in the k-th monitoring period, is the fourth sample refrigeration power of the jth sample ultra-low temperature freezer in the kth monitoring cycle, is the jth sample ultra-low temperature freezer in the kth monitoring cycle The fourth predicted sample temperature data at the moment, is the jth sample ultra-low temperature freezer in the kth monitoring cycle The cooling power of the fourth sample at the moment, n is the number of moments in the monitoring cycle, m is the number of monitoring cycles, N is the number of sample ultra-low temperature freezers, and is the preset weight, k≤m, j≤N, ,and , n, k, m, j and N are all positive integers.
6. The fault diagnosis method based on artificial intelligence according to claim 1, characterized in that: Obtaining a temperature matching score according to the first predicted temperature data, the second predicted temperature data, the first temperature data, and the second temperature data, including: According to the formula , Get temperature matching score ,in, is the first temperature data at the hth moment in the i-th monitoring cycle, is the first predicted temperature data at the hth moment in the i-th monitoring cycle, is the second temperature data at the hth moment in the i-th monitoring cycle, is the second predicted temperature data at the hth moment in the ith monitoring cycle, m is the number of monitoring cycles, h≤m, and both h and m are positive integers.
7. The fault diagnosis method based on artificial intelligence according to claim 1, characterized in that: When the first verification temperature data and the second verification temperature data meet preset requirements, determining a power matching score according to the first predicted cooling power, the second predicted cooling power, the first cooling power, and the second cooling power includes: According to the formula , Determining the Power Match Score ,in, Score the temperature match, is the first cooling power, is the first predicted cooling power, is the second cooling power, The second predicted cooling power.
8. An ultra-low temperature freezer, characterized in that: include: A primary compressor, a secondary compressor, an intermediate heat exchanger and a controller, wherein the primary compressor is used in a primary refrigeration cycle, the primary refrigeration cycle is used to transfer the heat of the intermediate heat exchanger to the air outside the ultra-low temperature freezer, and the secondary compressor is used in a secondary refrigeration cycle, the secondary refrigeration cycle is used to transfer the heat in the space inside the cabinet of the ultra-low temperature freezer to the intermediate heat exchanger; The controller is used to: At multiple moments in the i-th monitoring cycle, first temperature data in the intermediate heat exchanger and second temperature data in the cabinet of the ultra-low temperature freezer are respectively obtained, where i is a positive integer; Obtaining a first refrigeration power of the first-stage compressor and a second refrigeration power of the second-stage compressor in the i-th monitoring cycle; The first refrigeration power of the first-stage compressor in the i-th monitoring cycle, the second refrigeration power of the second-stage compressor, the first temperature data and the second temperature data at the first moment in the i-th monitoring cycle are input into the trained temperature prediction model to obtain the first predicted temperature data in the intermediate heat exchanger at multiple moments in the i-th monitoring cycle, and the second predicted temperature data in the cabinet of the ultra-low temperature freezer; Obtaining a temperature matching score according to the first predicted temperature data, the second predicted temperature data, the first temperature data, and the second temperature data; Input the first temperature data and the second temperature data at the first moment in the i-th monitoring cycle into the trained power prediction model to obtain the first predicted refrigeration power of the first-stage compressor and the second predicted refrigeration power of the second-stage compressor in the i-th monitoring cycle; Input the first predicted refrigeration power, the second predicted refrigeration power, the first temperature data at the first moment in the i-th monitoring cycle, and the second temperature data into the trained temperature prediction model to obtain the first verification temperature data in the intermediate heat exchanger at multiple moments in the i-th monitoring cycle, and the second verification temperature data in the cabinet of the ultra-low temperature freezer; In the case where the first verification temperature data and the second verification temperature data meet the preset requirements, the power matching score is determined according to the first predicted cooling power, the second predicted cooling power, the first cooling power, the second cooling power and the temperature matching score; wherein the preset requirements include that the first verification temperature data can remain stable relative to the first temperature data at the first moment, and that the second verification temperature data can maintain a sufficient rate of change relative to the second temperature data at the first moment; A fault score of the ultra-low temperature freezer is obtained according to the power matching score and the temperature matching score.
