Information processing device, method, and program
The information processing device improves prediction model accuracy by analyzing deviations and updating models based on operating conditions, effectively reducing prediction errors and enhancing abnormality detection in equipment.
Patent Information
- Application Number
- JP2023057927
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2023-03-31
- Publication Date
- 2024-11-20
- Estimated Expiration
- 2043-03-31
AI Technical Summary
Existing prediction models for equipment, such as air conditioners, often inaccurately predict whether the equipment is normal or abnormal, leading to prediction errors due to unknown conditions under which these errors occur.
An information processing device and method that includes a prediction model and a control unit to analyze deviations between predicted and actual values, updating the model based on operating conditions where deviations occur, and setting thresholds to improve accuracy by generating multiple models for each feature and determining abnormality based on deviation characteristics.
Enhances the accuracy of prediction models by reducing errors, allowing for precise detection of equipment abnormalities and reducing unnecessary maintenance visits.
Smart Images

Figure 0007587172000001 
Figure 0007587172000002 
Figure 0007587172000003
Abstract
Description
[Technical field]
[0001] The present disclosure relates to an information processing device, a method, and a program. [Background technology]
[0002] Conventionally, it is known to use a prediction model to obtain a predicted value from given data (Patent Document 1).
[0003] For example, in a predictive model that predicts values for equipment such as an air conditioner based on the operating data of the equipment, prediction errors may occur, such as predicting that the equipment is abnormal when it is actually normal, or predicting that the equipment is normal when it is actually abnormal. [Prior art documents] [Patent documents]
[0004] [Patent Document 1] JP 2014-059592 A Summary of the Invention [Problem to be solved by the invention]
[0005] However, it was not possible to improve the accuracy of the prediction model because it was not clear under what operating conditions the equipment would cause a large prediction error.
[0006] The present disclosure improves the accuracy of predictive models. [Means for solving the problem]
[0007] An information processing device according to a first aspect of the present disclosure includes: A prediction model that predicts a value related to the equipment from operation data of the equipment; and a control unit; The control unit is Based on information on whether a deviation value between a predicted value of a value related to the equipment obtained by inputting the operation data into the prediction model and an actual value of the value related to the equipment obtained by calculation from the operation data is within a normal range or an abnormal range of a predetermined deviation value, Analyzing the characteristics of operating conditions under which the deviation falls within an abnormal range; Based on the characteristics of the operating conditions under which the deviation falls within the abnormal range, the prediction model is updated using operating data under the operating conditions under which the deviation falls within the abnormal range and actual values related to the equipment.
[0008] According to the first aspect of the present disclosure, the accuracy of a prediction model can be improved.
[0009] A second aspect of the present disclosure is the information processing device according to the first aspect, The predictive model predicts a value when the value related to the device is normal; The control unit determines that the device is abnormal when the value relating to the deviation is large.
[0010] According to the second aspect of the present disclosure, it is possible to improve the accuracy of a predictive model generated by machine learning data when the device is normal.
[0011] A third aspect of the present disclosure is the information processing device according to the first aspect, the predictive model predicts values when the values related to the device are abnormal; The control unit determines that the device is abnormal when the value relating to the deviation is small.
[0012] According to the third aspect of the present disclosure, it is possible to improve the accuracy of a predictive model generated by machine learning data when an apparatus is abnormal.
[0013] A fourth aspect of the present disclosure is an information processing device according to any one of the first to third aspects, the normal range is a range in which the state of the device determined from the magnitude of the deviation value and the predetermined deviation value coincides with an actual state of the device, The abnormal range is a range in which the state of the device determined from the magnitude of the deviation value and the predetermined deviation value does not match the actual state of the device.
[0014] According to the fourth aspect of the present disclosure, the accuracy of the prediction model can be improved so as to eliminate an error between the predicted equipment state and the actual equipment state.
[0015] A fifth aspect of the present disclosure is an information processing device according to any one of the first to fourth aspects, The analysis of the characteristics involves sorting the operating conditions that fall within the abnormal range in order of frequency of occurrence and extracting a number of conditions from the top.
[0016] According to the fifth aspect of the present disclosure, the accuracy of a prediction model can be improved efficiently.
[0017] A sixth aspect of the present disclosure is an information processing device according to any one of the first to fifth aspects, The prediction model is updated by generating multiple models for each of the features and replacing the prediction model.
[0018] According to the sixth aspect of the present disclosure, prediction can be made according to the conditions under which the device is operating.
[0019] A seventh aspect of the present disclosure is an information processing device according to any one of the first to fifth aspects, The control unit sets a value by which the value relating to the deviation is to be subtracted or added for each of the characteristics.
[0020] According to the seventh aspect of the present disclosure, prediction can be made according to the conditions under which the device is operating.
[0021] An eighth aspect of the present disclosure is an information processing device according to any one of the first to fifth aspects, The control unit sets a threshold value for a deviation value used to determine an abnormality in the device for each of the characteristics.
[0022] According to the eighth aspect of the present disclosure, prediction can be made according to the conditions under which the device is operating.
[0023] A ninth aspect of the present disclosure is an information processing device according to any one of the first to eighth aspects, The equipment-related value is information relating to any one of the amount of refrigerant in the air conditioner, the degree of performance deterioration, and the degree of abnormality in the component parts.
[0024] According to the ninth aspect of the present disclosure, it is possible to improve the accuracy of a prediction model for predicting various values.
[0025] A method according to a tenth aspect of the present disclosure, comprising: A method executed by a control unit of an information processing device, A step of calculating information on whether a value relating to a deviation between a predicted value of a value relating to the equipment, obtained by inputting operation data of the equipment into a prediction model that predicts a value relating to the equipment from the operation data of the equipment, and an actual value of the value relating to the equipment, obtained by calculation from the operation data, is within a normal range or an abnormal range of a predetermined deviation value; Analyzing characteristics of operating conditions in which the deviation falls within an abnormal range; updating the prediction model using operating data and actual values related to the equipment when the operating conditions are such that the deviation falls within the abnormal range, based on characteristics of the operating conditions in which the deviation falls within the abnormal range; Includes.
[0026] A program according to an eleventh aspect of the present disclosure, a step of calculating information on whether a value relating to a deviation between a predicted value of a value relating to the equipment, obtained by inputting operation data of the equipment into a prediction model that predicts a value relating to the equipment from the operation data of the equipment, and an actual value of the value relating to the equipment, obtained by calculation from the operation data, is within a normal range or an abnormal range of a predetermined deviation value; A step of analyzing characteristics of operating conditions in which the deviation falls within an abnormal range; updating the prediction model using operating data and actual values related to the equipment when the operating conditions are such that the deviation falls within the abnormal range, based on characteristics of the operating conditions in which the deviation falls within the abnormal range; Execute the command. [Brief description of the drawings]
[0027] [Figure 1] FIG. 1 is a diagram showing an overall configuration according to an embodiment of the present disclosure. [Diagram 2] FIG. 1 is a diagram illustrating a hardware configuration of an information processing device according to an embodiment of the present disclosure. [Diagram 3] FIG. 2 is a diagram illustrating a functional configuration of a control unit according to an embodiment of the present disclosure. [Figure 4] FIG. 13 is a diagram for explaining a prediction model generated by machine learning of data when an apparatus according to an embodiment of the present disclosure is normal. [Diagram 5] FIG. 13 is a diagram for explaining a prediction model generated by machine learning of data when an apparatus according to an embodiment of the present disclosure is abnormal. [Figure 6] 13 is a diagram for explaining an evaluation of whether a deviation is within a normal range or an abnormal range according to an embodiment of the present disclosure. FIG. [Figure 7] 13 is a diagram for explaining an evaluation of whether a deviation is within a normal range or an abnormal range according to an embodiment of the present disclosure. FIG. [Figure 8] 10A and 10B are diagrams for explaining an analysis of characteristics of operating conditions in which the deviation falls within an abnormal range according to an embodiment of the present disclosure. [Figure 9]10A and 10B are diagrams for explaining an analysis of characteristics of operating conditions in which the deviation falls within an abnormal range according to an embodiment of the present disclosure. [Figure 10] FIG. 2 is a diagram for explaining generation of a prediction model for each characteristic of an operating condition according to an embodiment of the present disclosure. [Figure 11] 11 is a diagram for explaining setting of subtraction values and addition values for deviations for each characteristic of an operating condition and setting of a threshold value for the deviation for each characteristic of an operating condition according to an embodiment of the present disclosure. FIG. [Figure 12] 13 is a flowchart of a process for updating a prediction model based on characteristics of operating conditions in which a deviation falls within an abnormal range according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0028] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.
[0029] <Terminology explanation> "Operating data" refers to data (variables) acquired during the operation of equipment. For example, "operating data" is command values for equipment or components, status values of equipment or components (e.g., compressor speed, expansion valve opening, etc.), values indicating the environment of the equipment during operation (e.g., outdoor temperature, indoor temperature, outdoor humidity, indoor humidity, etc.), etc. For example, "operating data" is measured by sensors of the equipment or components (those that make up the equipment). "Equipment value" is any data related to equipment or components (things that make up the equipment). For example, an "equipment value" is data used to determine if there is an abnormality in the equipment, and indicates the status of the equipment or components. For example, an "equipment value" is information related to the amount of refrigerant in an air conditioner, the degree of performance degradation, or the degree of abnormality in a component. "Operating conditions" are the conditions under which the equipment is operating. In other words, "operating conditions" indicate the conditions under which the operating data was acquired.
[0030] <Overall composition> FIG. 1 is a diagram showing an overall configuration according to an embodiment of the present disclosure.
[0031] <<Information processing devices>> The information processing device 10 is a device that updates the prediction model by analyzing the characteristics of operating conditions (i.e., the conditions under which the operating data was obtained) under which the deviation between the predicted value of the value related to the equipment (more specifically, the value obtained by inputting the operating data of the equipment 11 into the prediction model) and the actual value of the value related to the equipment (more specifically, the value obtained by calculating from the operating data of the equipment 11) falls within an abnormal range.
[0032] The information processing device 10 is composed of one or more computers. The information processing device 10 acquires operation data of the device 11 (for example, receives the operation data of the device 11 from a communication device provided in the device 11 via an arbitrary network).
[0033] <<Equipment>> The devices 11 may be any devices such as an air conditioner, etc. Although three devices 11 are shown in Fig. 1, the number of devices 11 is not limited to this.
[0034] <Hardware configuration> 2 is a diagram showing a hardware configuration of an information processing device 10 according to an embodiment of the present disclosure. The information processing device 10 can include a control unit 1001, a main memory unit 1002, an auxiliary memory unit 1003, an input unit 1004, an output unit 1005, and an interface unit 1006. Each of these will be described below.
[0035] The control unit 1001 is a processor (eg, a central processing unit (CPU), a graphics processing unit (GPU), etc.) that executes various programs installed in the auxiliary storage unit 1003.
[0036] The main memory unit 1002 includes a non-volatile memory (Read Only Memory (ROM)) and a volatile memory (Random Access Memory (RAM)). The ROM stores various programs, data, etc. required for the control unit 1001 to execute various programs installed in the auxiliary memory unit 1003. The RAM provides a working area into which the various programs installed in the auxiliary memory unit 1003 are expanded when the control unit 1001 executes them.
[0037] The auxiliary storage unit 1003 is an auxiliary storage device that stores various programs and information used when the various programs are executed.
[0038] The input unit 1004 is an input device through which the operator of the information processing device 10 inputs various instructions to the information processing device 10 .
[0039] The output unit 1005 is an output device that outputs the internal state of the information processing device 10 and the like.
[0040] The interface unit 1006 is a communication device for connecting to a network and communicating with other devices.
[0041] <Functional configuration> 3 is a diagram showing a functional configuration of a control unit 1001 according to an embodiment of the present disclosure. The control unit 1001 can include an operating data acquisition unit 101, a prediction unit 102, a prediction model storage unit 103, an actual value calculation unit 104, a deviation evaluation unit 105, an analysis unit 106, a prediction model update unit 107, and a control value update unit 108. The control unit 1001 can function as the operating data acquisition unit 101, the prediction unit 102, the actual value calculation unit 104, the deviation evaluation unit 105, the analysis unit 106, the prediction model update unit 107, and the control value update unit 108 by executing a program.
[0042] The operation data acquisition unit 101 acquires operation data of the device 11 .
[0043] The prediction unit 102 inputs the operation data of the device 11 acquired by the operation data acquisition unit 101 into a prediction model stored in a prediction model storage unit 103, and outputs a predicted value related to the device.
[0044] The prediction model storage unit 103 stores a prediction model that outputs a value related to the device when operation data of the device 11 is input.
[0045] The actual value calculation unit 104 calculates an actual value of a value related to the device from the operation data of the device 11 acquired by the operation data acquisition unit 101.
[0046] The deviation evaluation unit 105 calculates information on whether a deviation value (e.g., the degree of deviation) between the predicted value of the value related to the equipment obtained by the prediction unit 102 and the actual value of the value related to the equipment obtained by the actual value calculation unit 104 is within a normal range or an abnormal range of a specified deviation value.
[0047] The analysis unit 106 analyzes the characteristics of the operating conditions under which the deviation falls within the abnormal range, based on information as to whether the deviation value calculated by the deviation evaluation unit 105 is within a normal range or an abnormal range of a predetermined deviation value magnitude.
[0048] The prediction model update unit 107 updates the prediction model using the operating data and the actual values related to the equipment when the operating conditions are such that the deviation falls within the abnormal range, based on the characteristics of the operating conditions in which the deviation falls within the abnormal range analyzed by the analysis unit 106. For example, the prediction model update unit 107 may update the prediction model by generating a prediction model for each characteristic of the operating conditions and replacing the existing prediction model (see [Generation of prediction models for each characteristic of the operating conditions] below), or may update the prediction model by relearning the existing prediction model (see [Relearning prediction model] below), or may update the prediction model by updating the explanatory variables of the prediction model (see [Updating explanatory variables of prediction model] below).
[0049] [Generating predictive models for each driving condition feature] The prediction model update unit 107 generates a prediction model for each characteristic of an operating condition where the deviation falls within an abnormal range (i.e., generates each prediction model by machine learning the operating data and equipment values for each operating condition as learning data), and replaces existing prediction models with multiple prediction models generated for each characteristic of the operating condition.
[0050] [Retraining the predictive model] The prediction model update unit 107 can re-learn the existing prediction model by using the actual values of the operating data and the equipment values under operating conditions where the deviation is in the abnormal range as learning data (for example, by adding them to the existing learning data).
[0051] [Update explanatory variables for forecast model] The prediction model update unit 107 can use variables (i.e., operating data) related to operating conditions in which the deviation falls within the abnormal range as explanatory variables of the prediction model (for example, by replacing existing explanatory variables, adding to existing explanatory variables, etc.).
[0052] The control value update unit 108 can perform the following processing when determining an abnormality in the device 11 using a prediction model, based on the results of the analysis by the analysis unit 106.
[0053] [Setting the subtraction or addition value to the deviation] When determining an abnormality in the equipment 11 using a predictive model, the control value update unit 108 can set a value to be subtracted from or added to a value related to the deviation (e.g., the deviation degree) for each characteristic of the operating conditions that causes the deviation analyzed by the analysis unit 106 to fall within the abnormal range.
[0054] [Setting deviation threshold] When determining whether or not an abnormality exists in the equipment 11 using a predictive model, the control value update unit 108 can set a threshold value for the deviation value (e.g., the degree of deviation) used to determine whether or not an abnormality exists in the equipment 11 for each characteristic of the operating conditions for which the deviation analyzed by the analysis unit 106 falls within the abnormal range.
[0055] <<Prediction model>> The prediction model will be described in detail below with reference to Figures 4 and 5. In this specification, the case where it is determined whether the device 11 is normal or abnormal will be mainly described, but it may be determined whether the device 11 is normal, has an abnormality degree of 1, has an abnormality degree of 2, ..., etc.
[0056] 4 is a diagram for explaining a prediction model generated by machine learning of data when the device 11 according to an embodiment of the present disclosure is normal. The prediction model is generated by machine learning of operation data and values related to the device when the device 11 is normal (more specifically, values related to the device determined by a service engineer or the like after checking the state of the device 11 on-site) as learning data. In other words, the prediction model is trained so that the error between the values related to the device output when the operation data of the device 11 is input to the prediction model and the values related to the device that are learning data is reduced. When the operation data of the device 11 is input to the prediction model, the values related to the device when the device 11 is normal are output.
[0057] In [Equipment Values] in Fig. 4, predicted values and actual values of the equipment values (i.e., equipment values calculated from operation data) at each time are shown. A normal period is when the equipment 11 is normal, and an abnormal period is when the equipment 11 is abnormal. As shown in [Equipment Values] in Fig. 4, a deviation between the actual value and the predicted value should occur in the abnormal period.
[0058] [Deviation] in Fig. 4 shows the deviation between the actual value and the predicted value at each time. There should be no deviation during a normal period, but deviation should occur during an abnormal period. In determining whether the device 11 is abnormal, a value related to such deviation (for example, the degree of deviation (the degree of deviation)) is compared with a threshold value, and if the value related to the deviation exceeds the threshold value, the device 11 is determined to be abnormal.
[0059] FIG. 5 is a diagram for explaining a prediction model generated by machine learning data when the device 11 according to an embodiment of the present disclosure is abnormal. The prediction model is generated by machine learning the operation data and the values related to the device (more specifically, the values related to the device determined by a service engineer or the like checking the state of the device 11 on-site) when the device 11 is abnormal (for example, broken) as learning data. That is, the prediction model is trained so that the error between the values related to the device output when the operation data of the device 11 is input to the prediction model and the values related to the device that are learning data is reduced. When the operation data of the device 11 is input to the prediction model, the values related to the device when the device 11 is abnormal are output.
[0060] In [Equipment Values] in Fig. 5, predicted values and actual values of the equipment values (i.e., equipment values calculated from operation data) at each time are shown. A normal period is when the equipment 11 is normal, and an abnormal period is when the equipment 11 is abnormal. As shown in [Equipment Values] in Fig. 5, there should be a deviation between the actual value and the predicted value during the normal period.
[0061] [Deviation] in Fig. 5 shows the deviation between the actual value and the predicted value at each time. Deviation does not occur during an abnormal period, but deviation should occur during a normal period. In determining whether the device 11 is abnormal, a value related to such deviation (for example, the degree of deviation (degree of deviation)) is compared with a threshold value, and if the value related to the deviation does not exceed the threshold value, the device 11 is determined to be abnormal.
[0062] <<Normal and Abnormal Ranges of Deviation>> The normal and abnormal ranges of deviation will be described in detail below with reference to FIGS.
[0063] 6 is a diagram for explaining an evaluation of whether the deviation is in the normal range or the abnormal range according to an embodiment of the present disclosure. Note that the case of a prediction model generated by machine learning data when the device 11 is normal will be explained.
[0064] The [Deviation] column in FIG. 6 shows the deviation between the actual equipment value (i.e., the equipment value calculated from the operation data) and the predicted value at each time.
[0065] [Device Status (Judgment)] in Fig. 6 shows the judgment result of the status of the device 11 at each time (i.e., whether the device 11 was judged to be normal or abnormal). In the case of Fig. 6, when the deviation between the actual value of the device (i.e., the value of the device calculated from the operation data) and the predicted value exceeds a threshold, the device 11 is judged to be abnormal.
[0066] [Device Status (Actual)] in FIG. 6 shows the actual status of the device 11 at each time (that is, whether the device 11 is normal or abnormal).
[0067] Comparing [Device Status (Determination)] and [Device Status (Actual)], in [Device Status (Actual)], the device 11 is normal until time t2 and becomes abnormal after time t2. However, in [Device Status (Determination)], the device 11 is determined to be abnormal not only after time t2 but also for a certain period of time from time t1.
[0068] In this way, the range in which the state of the device 11 determined by the deviation between the predicted value by the prediction model and the actual value does not match with the actual state of the device 11 is called an abnormal range (in this case, it is said that the value related to the deviation is in an abnormal range with a predetermined deviation value). Also, the range in which the state of the device 11 determined by the deviation between the predicted value by the prediction model and the actual value does not match with the actual state of the device 11 is called a normal range (in this case, it is said that the value related to the deviation is in a normal range with a predetermined deviation value).
[0069] It should be noted that, rather than evaluating whether or not the device 11 is in the abnormal range using the actual state of the device 11 as described above, it may be evaluated that the device is in the abnormal range when the value relating to the deviation (e.g., the degree of deviation) is a value outside a predetermined range (e.g., the degree of deviation is extremely large, etc.).
[0070] FIG. 7 is a diagram for explaining an evaluation of whether the deviation is within the normal range or the abnormal range according to an embodiment of the present disclosure.
[0071] As shown in FIG. 7, the operation data of the device 11 (in the example of FIG. 7, the outside air temperature, compressor speed, expansion valve opening, condensation temperature, and evaporation temperature) is acquired at each date and time. A value (e.g., deviation degree) regarding the deviation between the predicted value of the value related to the device and the actual value of the value related to the device when these operation data are input into the prediction model is calculated. This value regarding the deviation (e.g., deviation degree) is compared with a threshold value used to determine an abnormality of the device 11 to determine whether the device 11 is normal or abnormal. Then, depending on whether the determined state of the device 11 matches the actual state of the device 11, it is evaluated whether the deviation is within the normal range or the abnormal range. Using the operation data when the deviation is in the abnormal range (in the example of FIG. 7, 9:00 on October 18, 2021 and 10:00 on October 18, 2021), the characteristics of the operating conditions described below are analyzed.
[0072] <<Analysis of operating condition characteristics>> The analysis of the characteristics of the operating conditions will be described in detail below with reference to Figures 8 and 9. What operating conditions cause the deviation to fall within the abnormal range is analyzed.
[0073] FIG. 8 is a diagram for explaining an analysis of characteristics of operating conditions in which the deviation falls within an abnormal range according to an embodiment of the present disclosure.
[0074] For example, the analysis unit 106 performs cluster analysis using the operation data when the deviation is in the abnormal range. In detail, the analysis unit 106 performs cluster analysis based on the variables during operation of the device 11 (i.e., the operation data. In FIG. 8, variable 1 (e.g., compressor speed) and variable 2 (e.g., expansion valve opening)), and extracts operation conditions where the deviation frequently falls into the abnormal range (e.g., 70<compressor speed<90, 20<expansion valve opening<30).
[0075] The analysis unit 106 can sort the operating conditions in which the deviation falls within the abnormal range in order of frequency of occurrence (e.g., in order of the number of cluster samples) (e.g., operating condition D>operating condition B>operating condition A>operating condition C) and extract multiple operating conditions (e.g., operating condition D and operating condition B) from the top. In this case, the prediction model update unit 107 can exclude operating conditions with little data (operating condition A and operating condition C) and re-learn the prediction model using the operating data and actual values related to the equipment for operating condition D and operating condition B.
[0076] FIG. 9 is a diagram for explaining an analysis of characteristics of operating conditions in which the deviation falls within an abnormal range according to an embodiment of the present disclosure.
[0077] For example, the analysis unit 106 can perform analysis using not only the operation data when the deviation is in the abnormal range, but also the operation data when the deviation is in the normal range. In detail, the analysis unit 106 performs factor analysis on the operation data to extract the operation condition (e.g., 70<compressor speed, 20>expansion valve opening) that causes a difference between when the deviation is in the normal range and when the deviation is in the abnormal range.
[0078] <<Prediction models for each driving condition>> Fig. 10 is a diagram for explaining a prediction model for each characteristic of an operating condition according to an embodiment of the present disclosure. As shown in Fig. 10, when using a plurality of prediction models generated for each characteristic of an operating condition, the control value update unit 108 determines a prediction model to be used for prediction depending on which operating condition the operating data corresponds to (that is, a prediction model generated by machine learning using the operating data and values related to the equipment under the operating condition (corresponding to the operating data) as learning data), and the prediction unit 102 inputs the operating data into the prediction model determined by the control value update unit 108 and outputs a predicted value.
[0079] <<Setting subtraction and addition values for each characteristic of the operating conditions, and setting thresholds>> 11 is a diagram for explaining the setting of subtraction values and addition values for deviations for each characteristic of an operating condition and the setting of a threshold value for the deviation for each characteristic of an operating condition according to an embodiment of the present disclosure. Note that the case of a prediction model generated by machine learning data when the device 11 is normal will be explained.
[0080] 11 shows the deviation between the actual value of the equipment (i.e., the value of the equipment calculated from the operation data) and the predicted value at each time. For example, assume that the analysis of the characteristics of the operation conditions reveals that the deviation falls within the abnormal range when the operation conditions are A and B (for example, the equipment 11 is determined to be abnormal even though it is actually normal).
[0081] As shown in [Subtraction / Addition] in Figure 11, when operating condition A and operating condition B are met, a value related to the deviation used to determine an abnormality in the equipment 11 is calculated (for example, a value determined based on a representative value such as the average value or median value of the prediction error of the prediction model is subtracted or added) so that the equipment 11 is not determined to be abnormal even though it is normal, or is not determined to be normal even though it is abnormal.
[0082] As shown in [Threshold] in Figure 11, when operating conditions A and B are met, the threshold value of the deviation value used to determine an abnormality in equipment 11 is set to a threshold value at which equipment 11 is not judged to be abnormal even though it is normal, or a threshold value at which equipment 11 is not judged to be normal even though it is abnormal.
[0083] <Method> FIG. 12 is a flowchart of a process for updating a prediction model based on characteristics of operating conditions in which the deviation falls within an abnormal range according to an embodiment of the present disclosure.
[0084] In step 1 (S1), the control unit 1001 of the information processing device 10 acquires operation data of the device 11.
[0085] In step 2 (S2), the control unit 1001 of the information processing device 10 inputs the operation data of the device 11 acquired in S1 into a prediction model that predicts a value related to the device from the operation data of the device, and outputs a predicted value related to the device.
[0086] In step 3 (S3), the control unit 1001 of the information processing device 10 calculates actual values of the values related to the device from the operation data of the device 11 acquired in S1.
[0087] Note that S2 and S3 may be executed in the reverse order or simultaneously.
[0088] In step 4 (S4), the control unit 1001 of the information processing device 10 calculates information on whether the value relating to the deviation between the predicted value of the device value obtained in S2 and the actual value of the device value obtained in S3 is within a normal range or an abnormal range of the magnitude of the value relating to the deviation. If it is within the normal range, the following analysis is not performed.
[0089] In step 5 (S5), the control unit 1001 of the information processing device 10 analyzes the characteristics of the operating conditions under which the deviation falls within the abnormal range, based on information on whether the deviation value calculated in S4 is within the normal range or abnormal range of the magnitude of a specified deviation value.
[0090] In step 6 (S6), the control unit 1001 of the information processing device 10 updates the prediction model using the operating data and the actual values of the equipment at the operating conditions where the deviation falls within the abnormal range, based on the characteristics of the operating conditions where the deviation falls within the abnormal range analyzed in S5.
[0091] As described above, according to the embodiment of the present disclosure, an information processing device 10 can be provided, which includes a prediction model that predicts values related to equipment from operation data of the equipment, and a control unit 1001, and the control unit 1001 analyzes the characteristics of the operating conditions (i.e., under what conditions the operating data was acquired) in which the deviation is in the abnormal range based on information on whether the value related to the deviation (e.g., the degree of deviation) between the predicted value of the value related to the equipment obtained by inputting the operation data into the prediction model and the actual value of the value related to the equipment obtained by calculating from the operation data is in the normal range or abnormal range of the magnitude of the value related to the deviation, and updates the prediction model using the operating data and the actual value related to the equipment when the deviation is in the abnormal range based on the characteristics of the operating conditions in which the deviation is in the abnormal range. This makes it possible to improve the accuracy of the prediction model so that prediction errors do not occur. Therefore, it is possible to detect equipment failures in a wide range of operating conditions, and it is possible to reduce the number of service engineers who repair the equipment coming to work erroneously.
[0092] Preferably, the prediction model predicts values when the values related to the device are normal, and the control unit 1001 determines that the device 11 is abnormal when the deviation value is large. This makes it possible to improve the accuracy of the prediction model generated by machine learning data when the device 11 is normal.
[0093] Preferably, the prediction model predicts a value when the value related to the device is abnormal, and the control unit 1001 determines that the device 11 is abnormal when the deviation value is small. This makes it possible to improve the accuracy of the prediction model generated by machine learning data when the device 11 is abnormal.
[0094] Preferably, the normal range is a range in which the state of the device 11 determined from the value related to the deviation and the magnitude of the predetermined deviation value matches the actual state of the device 11, and the abnormal range is a range in which the state of the device 11 determined from the value related to the deviation and the magnitude of the predetermined deviation value does not match the actual state of the device. This makes it possible to improve the accuracy of the prediction model so as to eliminate an error between the predicted state of the device 11 and the actual state of the device 11.
[0095] Preferably, the analysis of the characteristics involves sorting the operating conditions that fall within the abnormal range in order of frequency of occurrence and extracting a plurality of the operating conditions that fall within the abnormal range in order of frequency of occurrence, thereby making it possible to efficiently improve the accuracy of the prediction model.
[0096] Preferably, the prediction model is updated by generating multiple models for each feature and replacing the prediction model, thereby making it possible to make predictions according to the conditions when the device 11 is operating.
[0097] Preferably, the control unit 1001 sets a value to be subtracted from or added to the value related to the deviation for each feature. This makes it possible to make a prediction according to the conditions when the device 11 is operating.
[0098] Preferably, the control unit 1001 sets a threshold value for the deviation value used for determining an abnormality in the device 11 for each feature. This makes it possible to make a prediction according to the conditions when the device 11 is operating.
[0099] Preferably, the value related to the equipment is information related to any one of the amount of refrigerant, the degree of performance deterioration, and the degree of abnormality of a component part of the air conditioner. This makes it possible to improve the accuracy of the prediction model for predicting various values.
[0100] According to an embodiment of the present disclosure, a method can be provided that is executed by the control unit 1001 of the information processing device 10, and includes a step of calculating information on whether a deviation value between a predicted value of a value related to the device obtained by inputting operation data of the device 11 into a prediction model that predicts a value related to the device from the operation data of the device and an actual value of the value related to the device obtained by calculating from the operation data is within a normal range or an abnormal range of a predetermined deviation value, a step of analyzing characteristics of an operation condition in which the deviation is within the abnormal range, and a step of updating the prediction model using the operation data and the actual value of the value related to the device when the deviation is within the abnormal range based on the characteristics of the operation condition in which the deviation is within the abnormal range. This makes it possible to improve the accuracy of the prediction model so as to prevent prediction errors.
[0101] According to an embodiment of the present disclosure, it is possible to provide a program that causes the control unit 1001 of the information processing device 10 to execute the following steps: a procedure for calculating information on whether a deviation value between a predicted value of a value related to the equipment obtained by inputting operation data of the equipment 11 into a prediction model that predicts a value related to the equipment from the operation data of the equipment and an actual value of the value related to the equipment obtained by calculation from the operation data is within a normal range or an abnormal range of a predetermined deviation value magnitude; a procedure for analyzing characteristics of an operation condition in which the deviation is within the abnormal range; and a procedure for updating the prediction model using the operation data and the actual value of the value related to the equipment when the deviation is within the abnormal range based on the characteristics of the operation condition in which the deviation is within the abnormal range. This makes it possible to improve the accuracy of the prediction model so as to prevent prediction errors.
[0102] Although the embodiments have been described above, it will be understood that various changes in form and details are possible without departing from the spirit and scope of the claims. [Explanation of symbols]
[0103] 10. Information processing device 11 Equipment 101 Operation data acquisition unit 102 Prediction Department 103 Prediction model memory unit 104 Actual value calculation unit 105 Deviation Evaluation Section 106 Analysis Department 107 Prediction model update unit 108 Control value update unit 1001 Control section 1002 Main memory 1003 Auxiliary storage unit 1004 Input section 1005 Output section 1006 Interface section
Claims
1. A prediction model that predicts a value related to the equipment from operation data of the equipment; and a control unit; The control unit is Based on information on whether a deviation value between a predicted value of a value related to the equipment obtained by inputting the operation data into the prediction model and an actual value of the value related to the equipment obtained by calculation from the operation data is within a normal range or an abnormal range of a predetermined deviation value, Analyzing the characteristics of operating conditions under which the deviation falls within an abnormal range; Based on characteristics of an operating condition in which the deviation falls within an abnormal range, the prediction model is updated using operating data in the operating condition in which the deviation falls within an abnormal range and an actual value related to the device. Information processing device.
2. The predictive model predicts a value when the value related to the device is normal; The information processing apparatus of claim 1 , wherein the control unit determines that the device is abnormal when the deviation value is large.
3. the predictive model predicts values when the values related to the device are abnormal; The information processing apparatus of claim 1 , wherein the control unit determines that the device is abnormal when the value relating to the deviation is small.
4. the normal range is a range in which the state of the device determined from the magnitude of the deviation value and the predetermined deviation value coincides with an actual state of the device, The information processing apparatus of claim 1 , wherein the abnormal range is a range in which a state of the device determined from a magnitude of the deviation value and the predetermined deviation value does not match an actual state of the device.
5. The information processing device according to claim 1 , wherein the analysis of the characteristics comprises sorting the operating conditions that fall within the abnormal range in order of frequency of occurrence and extracting a plurality of the operating conditions from the top.
6. The information processing device according to claim 1 , wherein the updating of the prediction model is performed by generating a plurality of models for each of the features and replacing the prediction model with the models.
7. The information processing device of claim 1 , wherein the control unit sets a value by which the value relating to the deviation is to be subtracted or added for each of the characteristics.
8. The information processing apparatus according to claim 1 , wherein the control unit sets a threshold value for a deviation value used to determine an abnormality in the device for each of the characteristics.
9. 2. The information processing apparatus according to claim 1, wherein the value relating to the equipment is information relating to any one of an amount of refrigerant, a performance deterioration state, and an abnormality state of a component part of an air conditioner.
10. A method executed by a control unit of an information processing device, A step of calculating information on whether a value relating to a deviation between a predicted value of a value relating to the equipment, obtained by inputting operation data of the equipment into a prediction model that predicts a value relating to the equipment from the operation data of the equipment, and an actual value of the value relating to the equipment, obtained by calculation from the operation data, is within a normal range or an abnormal range of a predetermined deviation value; Analyzing characteristics of operating conditions in which the deviation falls within an abnormal range; updating the prediction model using operating data and actual values related to the equipment when the operating conditions are such that the deviation falls within the abnormal range, based on characteristics of the operating conditions in which the deviation falls within the abnormal range; The method includes:
11. a step of calculating information on whether a value relating to a deviation between a predicted value of a value relating to the equipment, obtained by inputting operation data of the equipment into a prediction model that predicts a value relating to the equipment from the operation data of the equipment, and an actual value of the value relating to the equipment, obtained by calculation from the operation data, is within a normal range or an abnormal range of a predetermined deviation value; A step of analyzing characteristics of operating conditions in which the deviation falls within an abnormal range; updating the prediction model using operating data and actual values related to the equipment when the operating conditions are such that the deviation falls within the abnormal range, based on characteristics of the operating conditions in which the deviation falls within the abnormal range; A program for executing the above.
Citation Information
Patent Citations
Plant operation data prediction system and method
JP2010271852A
Facility state monitoring method and device of the same
JP2013025367A
Output value prediction device, output value prediction method and program for output value prediction method
JP2014059592A
Abnormality diagnosis device and abnormality diagnosis method
JP2020091561A
Apparatus management system
JP2020190877A
