A method, device, electronic device and storage medium for predicting optical module performance

Abstract

The embodiment of the present invention provides a method, device, electronic device and storage medium for predicting the performance of an optical module, the method comprising: obtaining a historical feature set of a target optical module whose performance is to be predicted; the historical feature set includes at least one first historical feature data recorded for the target optical module within a preset historical time period; inputting the historical feature set into a preset prediction model to obtain a target prediction feature set; the target order parameter in the preset prediction model is determined by iterative calculation using a particle swarm optimization algorithm PSO to a preset number of iterations; and generating a performance prediction result for the target optical module based on the target prediction feature set. Through the embodiment of the present invention, it is possible to establish a prediction model for the performance of an optical module by setting the number of iterations, which is suitable for automated scenarios and reduces the time and resources required to determine the order parameters in the prediction model, thereby improving the efficiency of predicting the performance of the optical module.

Description

Technical Field

[0001] The present invention relates to the technical field of detection, and in particular to a method, device, electronic device and storage medium for predicting the performance of an optical module. Background Art

[0002] As the external communication interface for various users and network facilities such as optical transmission, routers, and switches, optical modules are increasingly becoming an indispensable component of communication networks and user networks. Due to the diverse working scenarios and harsh working environment of optical modules, their operating status is affected by many factors. After long-term operation, their performance will gradually decline and "age". Aging will cause abnormalities and failures in related facilities, such as packet loss in user applications, bit errors in transmission systems, and unstable optical power. However, due to the multi-source and mixed characteristics of optical module quality data, it is difficult to quickly locate the optical module during troubleshooting, which will cause great trouble to maintenance personnel. It is difficult to conduct accurate troubleshooting.

[0003] Among the existing prediction methods such as feature recognition and classification intelligent algorithms, feature recognition has the defects of being difficult to select the number of decomposition layers and being easily affected by noise; classification intelligent algorithms such as neural networks are not suitable for scenarios such as optical module performance data that are highly dependent on time series, nonlinear, random, and sudden. In addition, large models such as DNN (Deep Neural Networks) + GRU (Gated Recurrent Unit) require separate processing of training and prediction, which has high prediction efficiency but low training efficiency.

[0004] Since the characteristic data of optical modules (for example, time and optical power) are all data with strong time series, time series analysis and prediction methods such as SARIMA (Seasonal Autoregressive Integrated Moving Average) can be used for prediction; however, the prediction effect of such methods is determined by the order parameters. Among them, the order parameters are usually determined in the following two ways:

[0005] Method 1: Use autocorrelation diagram and partial autocorrelation diagram to manually determine the order parameters. This method can find the optimal parameters, but it is highly subjective and not suitable for automation.

[0006] Method 2: Use the exhaustive algorithm + AIC (Akaike information criterion) value to determine the order parameters. This method can also find the optimal parameters and can be automated, but it takes too long to execute. Summary of the invention

[0007] In view of the above problems, a method, device, electronic device and storage medium for predicting the performance of an optical module are proposed to overcome the above problems or at least partially solve the above problems, including:

[0008] A method for predicting optical module performance, the method comprising:

[0009] Acquire a historical feature set of a target optical module whose performance is to be predicted; the historical feature set includes at least one first historical feature data recorded for the target optical module within a preset historical time period;

[0010] The historical feature set is input into a preset prediction model to obtain a target prediction feature set; the target order parameter in the preset prediction model is determined by iteratively calculating the particle swarm optimization algorithm PSO to a preset number of iterations;

[0011] A performance prediction result for the target optical module is generated according to the target prediction feature set.

[0012] Optionally, the method further comprises:

[0013] Acquire at least one set of training feature sets, wherein the at least one set of training feature sets includes at least one second historical feature data recorded for at least one optical module, wherein the at least one optical module includes at least the target optical module;

[0014] Obtaining a constraint range set for the order parameter, and using an exhaustive algorithm to search for a first order parameter corresponding to a minimum information criterion AIC value within the constraint range for a training feature set;

[0015] Using PSO to search for a second order parameter within the constraint range for the training feature set, and determining the corresponding number of iterations; the AIC value of the second order parameter is equal to the AIC value corresponding to the first order parameter;

[0016] The number of iterations with the largest value among the numbers of iterations corresponding to all training feature sets is used as the preset number of iterations;

[0017] The PSO is used to iteratively calculate the order parameters in the preset prediction model, and when the number of iterations reaches the preset number of iterations, the target order parameters are output.

[0018] Optionally, inputting the historical feature set into a preset prediction model to obtain a target prediction feature set includes:

[0019] Extracting a normal feature set and an abnormal feature set from the historical feature set; the first historical feature data in the normal feature set are all normal feature data, and the abnormal feature set includes at least one abnormal first historical feature data;

[0020] Inputting the normal feature set into a preset prediction model to obtain a first prediction feature set;

[0021] Inputting the abnormal feature set into a preset prediction model to obtain a second prediction feature set;

[0022] Generating a performance prediction result for the target optical module according to the target prediction feature set includes:

[0023] Acquire current characteristic data of the target optical module;

[0024] A performance prediction result for the target optical module is generated according to the current feature data, the first prediction feature set, and the second prediction feature set.

[0025] Optionally, the first prediction feature set and the second prediction feature set respectively include at least one prediction feature data, and generating a performance prediction result for the target optical module according to the current feature data, and the first prediction feature set and the second prediction feature set includes:

[0026] Determining a first standard deviation and a first mean value based on at least one predicted feature data in the first predicted feature set; and determining a second standard deviation and a second mean value based on at least one predicted feature data in the second predicted feature set;

[0027] Acquire current characteristic data of the target optical module, and determine a first absolute value of a difference between the current characteristic data and the first average value, and a first ratio of the first absolute value of the difference to the first standard deviation;

[0028] Determine a second absolute value of a difference between the abnormal feature data in the abnormal feature set and the second average value, and a second ratio of the second absolute value of the difference to the second standard deviation;

[0029] When the first ratio is greater than the second ratio, generating a performance prediction result indicating an abnormality;

[0030] When the first ratio is not greater than the second ratio, a performance prediction result indicating normality is generated.

[0031] Optionally, the historical feature set includes a plurality of first historical feature data arranged in chronological order, each first historical feature data corresponds to a time point; and extracting a normal feature set and an abnormal feature set from the historical feature set includes:

[0032] Determine a first time point in the historical feature set at which the first historical feature data is missing or has an erroneous value;

[0033] Determine a second time point that has normal first historical feature data recorded and is closest to the first time point;

[0034] Using the first historical feature data of the second time point to fill the first time point;

[0035] According to a preset time interval, the filled historical feature set is segmented to obtain at least one set of feature subsets;

[0036] If all the first historical feature data in a feature subset are normal feature data, the feature subset is used as a normal feature set;

[0037] If the last first historical feature data in a feature subset is abnormal feature data, the feature subset is used as an abnormal feature set.

[0038] Optionally, the method further comprises:

[0039] Outputting the performance prediction result;

[0040] receiving a determination result inputted for the performance prediction result;

[0041] When the determination result is correct, the normal feature set and the abnormal feature set are used as training feature sets.

[0042] An embodiment of the present invention further provides a device for predicting performance of an optical module, the device comprising:

[0043] An acquisition module, which acquires a historical feature set of a target optical module whose performance is to be predicted; the historical feature set includes at least one first historical feature data recorded for the target optical module within a preset historical time period;

[0044] The prediction module inputs the historical feature set into a preset prediction model to obtain a target prediction feature set; the target order parameter in the preset prediction model is determined by iteratively calculating the particle swarm optimization algorithm PSO to a preset number of iterations;

[0045] The performance prediction result generating module is used to generate a performance prediction result for the target optical module according to the target prediction feature set.

[0046] Optionally, the device further comprises:

[0047] A training module is used to obtain at least one set of training feature sets, wherein the at least one set of training feature sets includes at least one second historical feature data recorded for at least one optical module, and the at least one optical module at least includes the target optical module; obtain a constraint range set for the order parameter, and use an exhaustive algorithm to search for a first order parameter corresponding to a minimum information criterion AIC value within the constraint range for a training feature set; use PSO to search for a second order parameter within the constraint range for the training feature set, and determine the corresponding number of iterations; the AIC value of the second order parameter is equal to the AIC value corresponding to the first order parameter; among the numbers of iterations corresponding to all training feature sets, the number of iterations with the largest value is used as a preset number of iterations; use PSO to iteratively calculate the order parameters in the preset prediction model, and output the target order parameters when the number of iterations reaches the preset number of iterations.

[0048] Optionally, the prediction module includes:

[0049] An extraction submodule is used to extract a normal feature set and an abnormal feature set from the historical feature set; the first historical feature data in the normal feature set are all normal feature data, and the abnormal feature set includes at least one abnormal first historical feature data;

[0050] A first prediction submodule, configured to input the normal feature set into a preset prediction model to obtain a first prediction feature set;

[0051] A second prediction submodule, used for inputting the abnormal feature set into a preset prediction model to obtain a second prediction feature set;

[0052] The performance prediction result generating module is used to obtain current feature data of the target optical module; and generate a performance prediction result for the target optical module according to the current feature data, the first prediction feature set and the second prediction feature set.

[0053] Optionally, the first prediction feature set and the second prediction feature set respectively include at least one prediction feature data, and the performance prediction result generating module is used to determine a first standard deviation and a first average value based on at least one prediction feature data in the first prediction feature set; and, based on at least one prediction feature data in the second prediction feature set, determine a second standard deviation and a second average value; obtain current feature data of the target optical module, and determine a first absolute value of a difference between the current feature data and the first average value, and a first ratio of the first absolute value of the difference to the first standard deviation; determine a second absolute value of a difference between abnormal feature data in the abnormal feature set and the second average value, and a second ratio of the second absolute value of the difference to the second standard deviation; when the first ratio is greater than the second ratio, generate a performance prediction result indicating an abnormality; when the first ratio is not greater than the second ratio, generate a performance prediction result indicating a normality.

[0054] Optionally, the historical feature set includes multiple first historical feature data arranged in chronological order, each first historical feature data corresponds to a time point; the extraction submodule is used to determine the first time point in the historical feature set where the first historical feature data is missing or the first historical feature data is an erroneous value; determine the second time point that records normal first historical feature data and is closest to the first time point; use the first historical feature data of the second time point to fill the first time point; divide the filled historical feature set according to a preset time interval to obtain at least one set of feature subsets; if all the first historical feature data in a feature subset are normal feature data, then the feature subset is used as a normal feature set; if the last first historical feature data in a feature subset is abnormal feature data, then the feature subset is used as an abnormal feature set.

[0055] Optionally, the device further comprises:

[0056] The training set adding module is used to output the performance prediction result; receive the judgment result input for the performance prediction result; when the judgment result is correct, use the normal feature set and the abnormal feature set as the training feature set.

[0057] An embodiment of the present invention further provides an electronic device, including a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program implements the above optical module performance prediction method when executed by the processor.

[0058] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above optical module performance prediction method is implemented.

[0059] The embodiments of the present invention have the following advantages:

[0060] In an embodiment of the present invention, a historical feature set of a target optical module whose performance is to be predicted is first obtained; the historical feature set includes at least one first historical feature data recorded for the target optical module within a preset historical time period; the historical feature set is input into a preset prediction model to obtain a target prediction feature set; the target order parameter in the preset prediction model is determined by iterative calculation using a particle swarm optimization algorithm PSO to a preset number of iterations; and a performance prediction result for the target optical module is generated based on the target prediction feature set. Through the embodiment of the present invention, a prediction model for optical module performance is established by setting the number of iterations, which is suitable for automated scenarios and reduces the time and resources required to determine the order parameters in the prediction model, thereby improving the efficiency of predicting optical module performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solution of the present invention, the accompanying drawings required for use in the description of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0062] Figure 1 is a flowchart of a method for predicting performance of an optical module according to an embodiment of the present invention;

[0063] Figure 2 is a flowchart of another method for predicting performance of an optical module according to an embodiment of the present invention;

[0064] Figure 3 is a schematic diagram of a prediction scenario of optical module performance according to an embodiment of the present invention;

[0065] Figure 4 is a flowchart of steps for determining target order parameters according to an embodiment of the present invention;

[0066] Figure 5 is a schematic diagram of a prediction process of optical module performance according to an embodiment of the present invention;

[0067] Figure 6 The present invention is a block diagram of a device for predicting the performance of an optical module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0068] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all 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.

[0069] Reference Figure 1 , shows a flowchart of a method for predicting the performance of an optical module according to an embodiment of the present invention, which may include the following steps:

[0070] Step 101: Acquire a historical feature set of a target optical module whose performance is to be predicted; the historical feature set includes at least one first historical feature data recorded for the target optical module within a preset historical time period.

[0071] The target optical module may refer to an optical module whose performance is to be predicted, which may be an external communication interface deployed in optical transmission, routers, switches, and the like.

[0072] The first historical characteristic data may refer to characteristic data obtained after recording the performance of the target optical module at a preset historical time point, for example, the optical power of the target optical module; the historical characteristic set may be composed of one or more first historical characteristic data.

[0073] When the performance of the target optical module needs to be predicted, a historical feature set of the target optical module may be first obtained to predict the future performance trend of the target optical module based on the historical feature set of the target optical module.

[0074] Step 102: Input the historical feature set into the preset prediction model to obtain the target prediction feature set; the target order parameter in the preset prediction model is determined by iteratively calculating the particle swarm optimization algorithm PSO to a preset number of iterations.

[0075] The target order parameters in the preset prediction model may be iteratively calculated using a PSO (Particle Swarm Optimization) algorithm.

[0076] In practical applications, if the order parameters are manually determined using autocorrelation graphs and partial autocorrelation graphs, it may not be suitable for automated scenarios; and if the order parameters are determined using an exhaustive algorithm + AIC value, it may consume too much time and resources. Therefore, the embodiment of the present invention may first set a preset number of iterations; then, based on the preset number of iterations, the target order parameters in the preset prediction model are determined, so that the preset prediction model can be adapted to automated scenarios without consuming too much time and resources due to long-term iterations.

[0077] Specifically, when PSO is used to iteratively calculate the target order parameters, when the number of iterations reaches a preset number of iterations, the target order parameters are output; by setting the number of iterations to establish a prediction model for the performance of the optical module, the time and resources required to determine the order parameters in the prediction model are reduced, thereby improving the efficiency of predicting the performance of the optical module.

[0078] After establishing the preset prediction model, when the historical feature set of the target optical module is obtained, the historical feature set can be input into the preset prediction model; at this time, the preset prediction model can predict the feature data of the target optical module in a preset future time period based on the historical feature data, thereby obtaining the target prediction feature set.

[0079] The "preset future time period" may be set according to actual conditions, for example, three days or a week after the current time, etc., and the embodiment of the present invention does not impose any limitation on this.

[0080] Step 103: Generate a performance prediction result for a target optical module according to the target prediction feature set.

[0081] The performance prediction result may be used to indicate the future performance trend of the target optical module.

[0082] After obtaining the prediction result of the future feature data of the target optical module, the performance of the target optical module can be predicted based on the prediction result. Specifically, the performance prediction result of the target optical module can be generated based on the prediction feature data in the target prediction feature set.

[0083] In an embodiment of the present invention, a historical feature set of a target optical module whose performance is to be predicted is first obtained; the historical feature set includes at least one first historical feature data recorded for the target optical module within a preset historical time period; the historical feature set is input into a preset prediction model to obtain a target prediction feature set; the target order parameter in the preset prediction model is determined by iterative calculation using a particle swarm optimization algorithm PSO to a preset number of iterations; and a performance prediction result for the target optical module is generated based on the target prediction feature set. Through the embodiment of the present invention, a prediction model for optical module performance is established by setting the number of iterations, which is suitable for automated scenarios and reduces the time and resources required to determine the order parameters in the prediction model, thereby improving the efficiency of predicting optical module performance.

[0084] Reference Figure 2 , shows a flowchart of a method for predicting the performance of an optical module according to an embodiment of the present invention, which may include the following steps:

[0085] Step 201: Obtain a historical feature set of a target optical module whose performance is to be predicted.

[0086] When the performance of the target optical module needs to be predicted, a historical feature set of the target optical module may be first obtained to predict the future performance trend of the target optical module based on the historical feature set of the target optical module.

[0087] Step 202: extract a normal feature set and an abnormal feature set from the historical feature set; the first historical feature data in the normal feature set are all normal feature data, and the abnormal feature set includes at least one abnormal first historical feature data.

[0088] Specifically, one or more sets of normal feature sets may be extracted from the historical feature sets; all first historical feature data in the normal feature sets are normal feature data.

[0089] At the same time, one or more sets of abnormal feature sets can be extracted from the historical feature set; the abnormal feature set includes at least one abnormal first historical feature data.

[0090] It should be noted that whether the first historical characteristic data is normal or abnormal characteristic data can be determined based on pre-set judgment data; the pre-set judgment data can be set based on prior information, and the embodiment of the present invention does not limit this.

[0091] In one embodiment of the present invention, the historical feature set includes a plurality of first historical feature data arranged in chronological order, each first historical feature data corresponds to a time point; the normal feature set and the abnormal feature set can be extracted through the following sub-steps:

[0092] Sub-step 11: determining a first time point in the historical feature set at which the first historical feature data is missing or has an erroneous value.

[0093] First, the historical feature set can be preprocessed to ensure that the data used for prediction is stable.

[0094] Specifically, the first time point at which the first historical feature data is missing or is an erroneous value in the historical feature set can be determined; wherein the missing first historical feature data may mean that the first historical feature data should be recorded, but is not recorded; the first historical feature data being an erroneous value may mean that the recorded first historical feature data has a serious deviation, for example: exceeds a preset range.

[0095] The first time point may refer to a time point at which the first historical characteristic data should be recorded but is not recorded; or, it may refer to a time point at which the first historical characteristic data exceeding a preset range is recorded.

[0096] Sub-step 12: determining a second time point that has normal first historical feature data recorded and is closest to the first time point.

[0097] Then, the first time point may be filled with normal first historical characteristic data.

[0098] Specifically, a time point closest to the first time point may be determined first; and at the same time, it may be determined whether the first historical characteristic data recorded at the time point is normal or abnormal.

[0099] If the first historical characteristic data recorded at the time point is normal, the time point is used as the second time point.

[0100] If the first historical feature data recorded at the time point is abnormal, another time point closest to the first time point can be determined in addition to the time point; if the first historical feature data at the other time point is normal, the other time point is used as the second time point.

[0101] For example, if the first time point is the fourth day, then we can first determine whether the first historical feature data of the third and fifth days are normal; if the first historical feature data of the fifth day is normal and the first historical feature data of the third day is abnormal, then the fifth day is taken as the second time point.

[0102] As an example, a time point that is prior to the first time point, records normal first historical feature data, and is closest to the first time point can be used as the second time point; or a time point that is later than the first time point, records normal first historical feature data, and is closest to the first time point can be used as the second time point. The embodiments of the present invention are not limited to this.

[0103] Sub-step 13: Use the first historical feature data at the second time point to fill in the first time point.

[0104] After the second time point is determined, the first historical characteristic data corresponding to the second time point may be determined; and the first time point may be filled with the first historical characteristic data.

[0105] For example: the historical feature set is: (first day, W1), (second day, W2), (third day, W3), (fourth day, -), (fifth day, W5); if the fourth day is determined to be the first time point and the third day is the second time point, W3 can be filled to the second time point, thereby obtaining the historical feature set after filling: (first day, W1), (second day, W2), (third day, W3), (fourth day, W3), (fifth day, W5).

[0106] Sub-step 14: Segment the filled historical feature set at preset time intervals to obtain at least one set of feature subsets.

[0107] After the historical feature set is filled, the filled historical feature set can be segmented according to preset time intervals to obtain at least one set of feature subsets; for example: the filled historical feature set is (first day, W1), (second day, W2), (third day, W3), (fourth day, W3), (fifth day, W5), (sixth day, W6), and the preset time interval is 3 days, then feature subset 1 can be obtained: [(first day, W1), (second day, W2), (third day, W3)], feature subset 2: [(fourth day, W3), (fifth day, W5), (sixth day, W6)].

[0108] Sub-step 15: If all the first historical feature data in a feature subset are normal feature data, the feature subset is used as a normal feature set.

[0109] After obtaining the feature subsets, it can be determined whether each feature subset belongs to a normal feature set or an abnormal feature set.

[0110] Specifically, if all the first historical feature data in a feature subset are normal feature data, the feature subset may be used as a normal feature set.

[0111] Sub-step 16: If the last first historical feature data in a feature subset is abnormal feature data, the feature subset is used as an abnormal feature set.

[0112] If the last first historical feature data in a feature subset is abnormal feature data, the feature subset may be used as an abnormal feature set.

[0113] For example, the feature subset is: [(first day, W1), (second day, W2), (third day, W3)], where it is detected that W3 does not match the pre-set judgment data, then the feature subset can be used as an abnormal feature set.

[0114] In another embodiment of the present invention, duplicate data in the historical feature set may be eliminated first, and then a normal feature set and an abnormal feature set may be extracted from the historical feature set after the elimination process.

[0115] Step 203: Input the normal feature set into a preset prediction model to obtain a first prediction feature set.

[0116] After obtaining the normal feature set and the abnormal feature set, the normal feature set and the abnormal feature set may be respectively input into a preset prediction model for prediction.

[0117] Specifically, the normal feature set can be input into a preset prediction model; the preset prediction model can predict the trend of the future performance of the target optical module based on the first historical feature data in the normal feature set, thereby obtaining a first prediction feature set. The first prediction feature set may include feature data of multiple performances that may appear in the future for the target optical module, for example, it may include: predicting the solar power of the current day and predicting the solar power of the next day.

[0118] Step 204: Input the abnormal feature set into a preset prediction model to obtain a second prediction feature set.

[0119] At the same time, the abnormal feature set can be input into a preset prediction model; the preset prediction model can predict the future performance trend of the target optical module based on the first historical feature data in the abnormal feature set, thereby obtaining a second prediction feature set. The second prediction feature set and the first prediction feature set are prediction feature data obtained by prediction based on different historical feature sets.

[0120] Step 205: Acquire current characteristic data of the target optical module.

[0121] In the process of processing the historical feature set, you can also obtain the current feature data obtained by recording the target optical module at the current moment; the current feature data can refer to the feature data obtained by detecting the target optical module at the current time point, for example: the optical power of the target optical module at the current time point.

[0122] Step 206: Generate a performance prediction result for the target optical module according to the current feature data, the first prediction feature set and the second prediction feature set.

[0123] After obtaining the current feature data, and the first prediction feature set and the second prediction feature set, which are equivalent to the future, a performance prediction result for the target optical module can be generated based on the current feature data and the first prediction feature set and the second prediction feature set.

[0124] In an embodiment of the present invention, the first prediction feature set and the second prediction feature set may respectively include at least one prediction feature data; step 206 may include the following sub-steps:

[0125] Sub-step 21: determining a first standard deviation and a first mean value based on at least one prediction feature data in the first prediction feature set; and determining a second standard deviation and a second mean value based on at least one prediction feature data in the second prediction feature set.

[0126] After obtaining the first prediction feature set and the second prediction feature set, a performance prediction result for the target optical module may be generated based on the first prediction feature set and the second prediction feature set.

[0127] Specifically, the standard deviation and the average value of at least one prediction feature data in the first prediction feature set may be calculated; for the sake of distinction, the standard deviation is taken as the first standard deviation, and the average value is taken as the first average value.

[0128] As an example, the first standard deviation can be calculated by the following formula:

[0129]

[0130] Here, σ can refer to the first standard deviation, x i It may refer to any prediction feature data in the first prediction feature set, x may refer to the first average value, and n may refer to the number of prediction feature data included in the first prediction feature set.

[0131] As another example, the first prediction feature set may include feature data for the prediction day, feature data for the prediction day 2, and feature data for the prediction day 3; the first average value may be calculated by the following formula:

[0132] The first average value=(the feature data of the predicted day+the feature data of the predicted second day+the feature data of the predicted third day) / 3.

[0133] At the same time, the standard deviation and the average value of at least one prediction feature data in the second prediction feature set may be calculated; for the sake of distinction, the standard deviation is used as the second standard deviation, and the average value is used as the second average value.

[0134] Sub-step 22: determining a first absolute value of a difference between the current characteristic data and the first average value, and a first ratio of the first absolute value of the difference to the first standard deviation.

[0135] After obtaining the first standard deviation, the first average value, the second standard deviation, the second average value, and the current characteristic data, the performance prediction result for the target optical module can be generated by combining the numerical relationship between the first standard deviation, the first average value, the second standard deviation, the second average value, and the current characteristic data.

[0136] Specifically, a first absolute value of a difference between the current characteristic data and a first average value may be calculated first; and a first ratio of the first absolute value of the difference to a first standard deviation may be calculated.

[0137] Sub-step 23: determining a second absolute value of a difference between the abnormal feature data in the abnormal feature set and the second average value, and a second ratio of the second absolute value of the difference to the second standard deviation.

[0138] At the same time, abnormal feature data can be extracted from the abnormal feature set, and a second absolute value of the difference between the abnormal feature data and the second average value can be calculated; and a second ratio of the second absolute value of the difference to the second standard deviation can be calculated.

[0139] As an example, the abnormal feature set may include multiple groups; each group may correspond to a second ratio; in this case, a preset value can be set, and the preset value and all the second ratios can be taken as a set; then the minimum value in the set is used as the second ratio to be compared with the first ratio in the subsequent step 207.

[0140] For example: the abnormal feature set includes two groups (Group 1 and Group 2); the second ratio of Group 1 is 2.72, and the second ratio of Group 2 is 3.59; the preset value is 3; then the second ratio compared with the first ratio in the subsequent step 207 = min(min(2.72,3.59), 3) = 2.72.

[0141] Sub-step 24: When the first ratio is greater than the second ratio, generating a performance prediction result indicating an abnormality.

[0142] After the first ratio and the second ratio are obtained, the magnitude relationship between the first ratio and the second ratio may be compared.

[0143] If the first ratio is greater than the second ratio, it may indicate that the difference between the current feature data and the first average value of the first prediction set obtained based on the normal feature set prediction is large; at this time, it may be determined that the target optical module may have an abnormality; furthermore, a performance prediction result indicating that the target optical module may have an abnormality may be generated.

[0144] Sub-step 25: When the first ratio is not greater than the second ratio, a performance prediction result indicating normality is generated.

[0145] If the first ratio is not greater than the second ratio, it may indicate that the difference between the current feature data and the first average value of the first prediction set obtained based on the normal feature set prediction is small; at this time, it may be determined that the target optical module may not be abnormal; furthermore, a performance prediction result indicating that the target optical module is normal may be generated.

[0146] For example: the preset number of iterations is 8, the minimum AIC value is -104.57, the execution time is about 1.2 seconds; the abnormal feature set includes two groups.

[0147] Obtaining predicted values: Use the target order parameters to predict the feature data for the next 10 days based on the feature data in the historical feature set.

[0148] Model prediction value acquisition: The standard deviation of SARIMA statistics is 0.1323, and the first average value of 17.57 is obtained by taking the data of the last 3 days.

[0149] Ratio calculation:

[0150] a) Input the two sets of abnormal feature sets into SARIMA one by one to obtain the results. The second mean value of group 1 is 17.57, the abnormal feature data is 17.93, and the second standard deviation is 0.1323; the second mean value of group 2 is 17.67, the abnormal feature data is 18.18, and the second standard deviation is 0.1418.

[0151] b) The second ratio = (|second mean value - abnormal characteristic data) / second standard deviation. The second ratio of group 1 is 2.72, and the second ratio of group 2 is 3.59.

[0152] c) Second ratio compared with the first ratio = min(min(all second ratios), 3) = 2.72.

[0153] Among them, data outside of 3 standard deviations can be set as abnormal.

[0154] d) First ratio = (|first average value - current characteristic data) / first standard deviation = 0.2268.

[0155] Alarm status calculation: the first ratio is less than the second ratio. The output result is S00, where S00 represents a normal performance prediction result.

[0156] Step 207: Output the performance prediction result.

[0157] After the performance prediction result for the target optical module is generated, the performance prediction result may be output so that an administrator may judge the accuracy of the performance prediction result based on the performance prediction result and the actual performance result.

[0158] Step 208: Receive a determination result inputted for the performance prediction result.

[0159] Specifically, the administrator can obtain the actual performance result of the target optical module based on the actual current performance of the target optical module; then, the administrator can compare the actual performance result with the performance prediction result to determine the accuracy of the performance result.

[0160] If the administrator determines that the actual performance result is consistent with the performance prediction result, the administrator may input a determination result indicating that it is correct.

[0161] If the administrator determines that the actual performance results are inconsistent with the performance prediction results, the administrator may enter a determination result indicating an error.

[0162] For example, the performance prediction results and the actual performance results can be displayed on the administrator's device; then, the administrator can input the corresponding determination results based on subjective judgment.

[0163] Step 209: When the judgment result is correct, the normal feature set and the abnormal feature set are used as training feature sets.

[0164] If the received judgment result is a correct judgment result, it can be said that the prediction process of the target optical module performance based on the historical feature set is accurate; at this time, the normal feature set and the abnormal feature set can be automatically used as training feature sets to increase the training samples of the model, so that the prediction ability of the model can be maintained stable.

[0165] like Figure 3 When the performance of the target optical module 310 needs to be predicted, the prediction system 320 deployed with the method of steps 201 to 211 can be used to collect historical feature data from the target optical module 310 and predict the performance of the target optical module 310 based on the historical feature set.

[0166] At the same time, the prediction system 320 can send the performance prediction results to the administrator device 330, so that the administrator device 330 can determine the accuracy of the results output by the prediction system based on the performance prediction results and the actual performance results; if the determination result is that the performance prediction results are correct, an instruction can be sent to the prediction system 320 to use the historical feature set as the training feature set.

[0167] In an embodiment of the present invention, a historical feature set of a target optical module whose performance is to be predicted is obtained; a normal feature set and an abnormal feature set are extracted from the historical feature set; the first historical feature data in the normal feature set are all normal feature data, and the abnormal feature set includes at least one abnormal first historical feature data; the normal feature set is input into a preset prediction model to obtain a first prediction feature set; the abnormal feature set is input into a preset prediction model to obtain a second prediction feature set; the current feature data of the target optical module is obtained; a performance prediction result for the target optical module is generated based on the current feature data, as well as the first prediction feature set and the second prediction feature set; the performance prediction result is output; a judgment result input for the performance prediction result is received; when the judgment result is correct, the normal feature set and the abnormal feature set are used as training feature sets. Through the embodiment of the present invention, it is realized to establish a prediction model for the performance of the optical module by setting the number of iterations, which is suitable for automated scenarios and reduces the time and resources consumed by determining the order parameters in the prediction model, thereby improving the efficiency of predicting the performance of the optical module.

[0168] In addition, by outputting the performance prediction results and receiving the judgment results input for the performance prediction results; when the judgment results are correct, the normal feature set and the abnormal feature set are used as the training feature set. By setting up an alarm feedback mechanism, the training samples of the model can be automatically increased, so that the prediction ability of the model can be maintained stable.

[0169] The following is a detailed description of the process of determining the target order parameters:

[0170] Reference Figure 4 , shows a flowchart of steps for determining target order parameters according to an embodiment of the present invention, which may include the following steps:

[0171] Step 401: Acquire at least one set of training feature sets, where the at least one set of training feature sets includes at least one second historical feature data recorded for at least one optical module, and the at least one optical module includes at least a target optical module.

[0172] The training feature set may include at least one second historical feature data recorded for at least one optical module, and the at least one optical module includes at least a target optical module and may also include optical modules of other signals, which is not limited in the embodiment of the present invention.

[0173] The second historical characteristic data may refer to characteristic data obtained after recording the performance of any optical module of the at least one optical module at a preset historical time point.

[0174] When it is necessary to establish a preset prediction model for an optical module device such as a target optical module, at least one set of training feature sets recorded for at least one optical module may be first obtained.

[0175] Step 402: Obtain a constraint range set for the order parameter, and use an exhaustive algorithm to search for a first order parameter corresponding to a minimum information criterion AIC value within the constraint range for a training feature set.

[0176] The constraint range may be set according to actual conditions, for example, between 0 and 3, and the embodiment of the present invention does not limit this.

[0177] After obtaining at least one set of training feature sets, the constraint range set for the order determination parameters can be obtained.

[0178] Then, a training feature set can be selected from at least one set of training feature sets, and an exhaustive algorithm is used to search for the order parameter corresponding to the minimum AIC (Akaike information criterion) value within the constraint range; for ease of distinction, the order parameter is used as the first order parameter.

[0179] Step 403: Use PSO to search for a second order parameter within a constraint range for a training feature set, and determine the corresponding number of iterations; the AIC value of the second order parameter is equal to the AIC value corresponding to the first order parameter.

[0180] In practical applications, if an exhaustive algorithm is used to determine the order parameters, although the optimal parameters can be obtained, its execution takes too long and consumes a lot of time and resource costs.

[0181] In order to adapt to the automation scenario and reduce the consumption of time and resource costs, the embodiment of the present invention can use PSO (Particle Swarm Optimization) to search for the order parameters within the constraint range for a training feature set.

[0182] When searching, PSO will find the optimal solution through iteration; at this time, the AIC value corresponding to the first order parameter can be used as the optimal solution; when the AIC value of the order parameter searched by PSO is equal to the AIC value of the first order parameter, the order parameter searched by PSO can be used as the second order parameter.

[0183] As an example, PSO is initialized as a group of random particles (random solutions). The optimal solution can be found through iteration. During the iteration, the particles are updated by tracking pbest and gbest. After finding these two optimal values, their positions and velocities can be updated by the following formula.

[0184] Position during iteration (formula):

[0185] in, represents the position at time t, p t-1 represents the position at time t-1, V i t-1 represents the speed at time t-1, and i represents the number of samples (i.e. the number of feature data).

[0186] Speed ​​during iteration (formula):

[0187]

[0188] Among them, P1 part belongs to the inertia part, w represents the weight factor corresponding to inertia; P2 part represents the individual knowledge part, C1 represents the learning factor, which can be set to 2, R1 represents a random number from 0 to 1, pbest represents the best result learned by the individual, p t-1 represents the parameter vector at time t-1; P3 represents the population knowledge part, C2 represents the learning factor, which can be set to 2, R2 represents a random parameter of 0 to 1, gbest represents the best result learned by the population, and pt-1 represents the parameter vector at time t-1).

[0189] In practical applications, V i t is a non-integer, R1 and R2 are random constants, for V i t Perform rounding operations.

[0190] In practical applications, for each training feature set, the number of iterations for searching the second order parameter using PSO can be determined separately.

[0191] Step 404: Among the iteration numbers corresponding to all training feature sets, the iteration number with the largest value is used as the preset iteration number.

[0192] Then, the number of iterations with the largest value among the numbers of iterations corresponding to all training feature sets can be used as the preset number of iterations for determining the target order parameter of the preset prediction model.

[0193] For example:

[0194] The preset prediction model is SARIMA; the training feature set includes two groups; an exhaustive algorithm is used to exhaustively enumerate all possible order parameters for each group (the p, q, d, P, Q, D constraints are set between 0 and 3), and finally all AIC values ​​are obtained. The AIC values ​​of group 1 are shown in Table 1, and the AIC values ​​of group 2 are shown in Table 2.

[0195] Table 1:

[0196] SARIMA AIC value SARIMA(0,1,1)(0,0,0,2) -104.57 SARIMA(1,1,1)(0,0,0,12) -102.88 SARIMA(0,1,2)(0,0,0,12) -101.49 SARIMA(2,1,1)(0,0,0,12) -100.02 SARIMA(1,1,1)(0,2,0,12) 45.56 SARIMA(1,1,2)(0,2,0,12) 45.68 SARIMA(1,0,2)(2,2,1,12) 1559.48

[0197] Table 2:

[0198] SARIMA AIC value SARIMA(0,1,1)(0,0,0,12) -65.45 SARIMA(1,1,1)(1,0,0,12) -60.33 SARIMA(2,0,1)(1,0,0,12) -59.29 SARIMA(2,1,1)(1,0,0,12) -56.08 SARIMA(1,0,1)(0,2,0,12) 65.28 SARIMA(1,1,2)(0,2,0,12) 100.68 SARIMA(1,0,2)(1,2,1,12) 156.25

[0199] The best result of group 1 is SARIMA(0,1,1)(0,0,0,12), with an AIC value of -104.57 and an overall execution time of about 20 minutes. The best result of group 2 is SARIMA(0,1,1)(0,0,0,12), with an AIC value of -65.45 and an overall execution time of about 23 minutes.

[0200] Use PSO to find the order parameters of the SARIMA module (p, q, d, P, Q, D constraints are set to 0 to 3). If the AIC output by the SARIMA module in group 1 is equal to -104.57 (the minimum AIC value obtained by exhaustive calculation), the PSO search ends and the number of iterations is recorded as 8. The execution time is about 37 seconds.

[0201] If the AIC output by the SARIMA module of group 2 is equal to -65.45 (the minimum AIC value obtained by exhaustive calculation), the PSO search ends, and the number of iterations is recorded as 6, and the execution time is about 23 seconds. The maximum value is taken as the preset number of iterations 8.

[0202] In practical applications, the preset number of iterations may be calculated only once before the target optical module is deployed; when a new optical module needs to be predicted, it may be recalculated based on the characteristic data of the new optical module, which is not limited in the embodiments of the present invention.

[0203] Step 405: PSO is used to iteratively calculate the order parameters in the preset prediction model, and when the number of iterations reaches the preset number of iterations, the target order parameters are output.

[0204] As an example, the preset prediction model may refer to SARIMA.

[0205] Among them, SARIMA can be expressed as follows:

[0206]

[0207] Among them, y t represents the value at time t (which can refer to the characteristic data of the target optical module at time t), b uses a constant offset, γ i represents the autocorrelation coefficient, θ i ,φ i , τ i represents the model parameters, ε represents zero-mean white noise, s represents the seasonal parameter, and i represents the number of samples (i.e., the number of feature data).

[0208] The P1 part represents autoregression, which calculates the relationship between the current value and the historical value. The p-order autoregression means that the current value is related to the previous p values.

[0209] The P2 part represents the moving average model, which mainly eliminates the random fluctuations of the data. The q-order autoregression means that the current value is related to the previous q errors.

[0210] The P3 part represents the seasonal partial autoregression, which calculates the relationship between the current value and the historical value. The P-order autoregression means that the current value is related to the previous ts*P values.

[0211] The P4 part represents the seasonal partial moving average model, and the Q-order autoregression means that the current value is related to the previous ts*Q errors.

[0212] The non-seasonal parameters (p, d, q) and seasonal parameters (P, D, Q)s are called the order parameters in SARIMA. For the same training data, SARIMA uses different order parameters, and its data fitting ability is different. Under normal circumstances, the AIC value is used to evaluate the quality of the order parameters. The smaller the AIC value, the more appropriate the parameter is, and the better the data fitting is.

[0213] If the input data is not stable, d-order difference processing is required. If the data has seasonal changes and is not stable, D-order difference processing is required to adjust the data to a stable state.

[0214] For example, after preprocessing, the optical power data of the optical module is stable, and the standard deviation fluctuates within a small range. The data is differentially processed and the stability test is performed through the Ljung-Box test. The test results can be seen in Table 3. P-value < 0.05 proves that the original sequence is not white noise, and time series modeling can continue.

[0215] Table 3:

[0216] Test items value Statistics 26.68648 P-value 0.00023

[0217] After determining the preset number of iterations, the PSO iteration can be used to calculate the order parameters in the preset prediction model.

[0218] At the same time, the number of current iterations can be counted; when the number of iterations reaches the preset number of iterations, the loop is exited and the target order parameters are output; when the number of iterations has not yet reached the preset number of iterations, the order parameters in the preset prediction model are continued to be iteratively calculated.

[0219] For example:

[0220] ① Randomly initialize the PSO particle swarm velocity vector (p, q, d, P, Q, D) and position vector (p, q, d, P, Q, D, s). Among them, p is the number of autoregressive terms, d is the number of differences (order) made for the stationary series, and q is the number of sliding average terms; the variable constraints of seasonal parameters (P, Q, D) are the same as those of non-seasonal parameters, and s is the time interval.

[0221] ②. Input the initialization vector (p, q, d, P, Q, D, s) in ① and the power data corresponding to the optical module into SARIMA. After training, output the AIC value.

[0222] ③. Use the updated particle swarm position and velocity, and use the AIC value output by ② to update the population optimal value (pbest value and gbest value).

[0223] Determine whether the number of iterations reaches the upper limit. If so, exit the loop and output the target order parameter of SARIMA. If not, continue to search for the optimal solution.

[0224] like Figure 5 ,In the process of predicting the performance of the target optical module, it can include three parts: model ,optimal solution, anomaly detection and feedback processing.

[0225] The optimal solution part of the model:

[0226] Determine the AIC value: Initialize the constraint range of the SARIMA order parameters. For each set of preprocessed training feature sets, an exhaustive algorithm is used to search for the optimal SARIMA order parameters within the constraint range, and the minimum AIC value generated is recorded. For each set of preprocessed training feature sets, a PSO algorithm is used to search for the optimal SARIMA order parameters within the constraint range. When the AIC value generated by the PSO algorithm is equal to the minimum AIC value recorded by the exhaustive algorithm, the PSO exits the search early and records the number of iterations of the current group. Take the maximum number of iterations corresponding to all training feature sets as the preset number of iterations.

[0227] Whether the preset number of iterations is exceeded + target order parameters are determined: PSO is used to determine the target order parameters of SARIMA according to the preset number of iterations obtained.

[0228] Anomaly detection part:

[0229] Fitting the data: In the anomaly detection section, the normal feature set is input into SARIMA (fitting the data using the target order parameters described above).

[0230] Mean calculation:

[0231] a) The first average value = (the feature data of the predicted day + the feature data of the predicted second day + the feature data of the predicted third day) / 3.

[0232] b) First standard deviation.

[0233] Module standard deviation calculation: Under normal circumstances, data outside the 3 standard deviations are abnormal data. By adjusting the boundary division between abnormal data and normal data and re-dividing the boundary of abnormal data, better fitting can be achieved. The boundary is calculated in the following way:

[0234] a) Input all abnormal feature sets into SARIMA one by one to obtain all second means and second standard deviations.

[0235] b) Calculate the second ratio corresponding to each set of abnormal feature sets = (|second average value - abnormal feature data|) / second standard deviation; c) The second ratio compared with the first ratio = min(min(all second ratios), 3).

[0236] Alarm status calculation: If |first average value - current characteristic data|>second ratio*first ratio, the alarm status is set to S10 (emergency), otherwise the alarm status is set to S00 (normal); wherein S00 represents a normal performance prediction result, and S10 represents an abnormal performance prediction result.

[0237] Feedback processing part:

[0238] Alarm output: Output performance prediction results to the administrator's device.

[0239] Feedback result processing: Administrators can provide feedback based on performance prediction results and actual performance results, and input judgment results for the performance prediction results; then, based on the judgment results, it can be determined whether to use the normal feature set and the abnormal feature set as training feature sets to increase the training samples of the model.

[0240] Before the three steps, a training feature set may be obtained in advance and preprocessed. Specifically, the preprocessing may include: removing duplicates; filling missing values ​​and error values; and extracting a set including normal feature data and a combination of abnormal feature data.

[0241] In the embodiment of the present invention, the prediction model is established by setting the number of iterations, which reduces the time and resources consumed in determining the order parameters in the prediction model, thereby improving the efficiency of predicting the performance of the optical module.

[0242] Reference Figure 6 , shows a schematic diagram of the structure of a device for predicting the performance of an optical module according to an embodiment of the present invention, which may include the following modules:

[0243] An acquisition module 601, which acquires a historical feature set of a target optical module whose performance is to be predicted; the historical feature set includes at least one first historical feature data recorded for the target optical module within a preset historical time period;

[0244] Prediction module 602, inputs the historical feature set into a preset prediction model to obtain a target prediction feature set; the target order parameter in the preset prediction model is determined by iterative calculation using a particle swarm optimization algorithm PSO to a preset number of iterations;

[0245] The performance prediction result generating module 603 is used to generate a performance prediction result for a target optical module according to a target prediction feature set.

[0246] In an optional embodiment of the present invention, the device further comprises:

[0247] A training module is used to obtain at least one set of training feature sets, wherein the at least one set of training feature sets includes at least one second historical feature data recorded for at least one optical module, and the at least one optical module includes at least a target optical module; obtain a constraint range set for an order parameter, and use an exhaustive algorithm to search for a first order parameter corresponding to a minimum information criterion AIC value within the constraint range for a training feature set; use PSO to search for a second order parameter within the constraint range for a training feature set, and determine the corresponding number of iterations; the AIC value of the second order parameter is equal to the AIC value corresponding to the first order parameter; among the numbers of iterations corresponding to all training feature sets, the number of iterations with the largest value is used as a preset number of iterations; use PSO to iteratively calculate the order parameters in a preset prediction model, and output the target order parameters when the number of iterations reaches the preset number of iterations.

[0248] In an optional embodiment of the present invention, the prediction module 602 includes:

[0249] An extraction submodule is used to extract a normal feature set and an abnormal feature set from the historical feature set; the first historical feature data in the normal feature set are all normal feature data, and the abnormal feature set includes at least one abnormal first historical feature data;

[0250] A first prediction submodule, used for inputting a normal feature set into a preset prediction model to obtain a first prediction feature set;

[0251] A second prediction submodule, used for inputting the abnormal feature set into a preset prediction model to obtain a second prediction feature set;

[0252] The performance prediction result generating module 603 is used to obtain the current feature data of the target optical module; and generate the performance prediction result for the target optical module according to the current feature data, the first prediction feature set and the second prediction feature set.

[0253] In an optional embodiment of the present invention, the first prediction feature set and the second prediction feature set respectively include at least one prediction feature data, and the performance prediction result generation module 603 is used to determine a first standard deviation and a first average value based on at least one prediction feature data in the first prediction feature set; and, based on at least one prediction feature data in the second prediction feature set, determine a second standard deviation and a second average value; obtain current feature data of the target optical module, and determine a first absolute value of a difference between the current feature data and the first average value, and a first ratio of the first absolute value of the difference to the first standard deviation; determine a second absolute value of a difference between the abnormal feature data in the abnormal feature set and the second average value, and a second ratio of the second absolute value of the difference to the second standard deviation; when the first ratio is greater than the second ratio, generate a performance prediction result indicating an abnormality; when the first ratio is not greater than the second ratio, generate a performance prediction result indicating a normality.

[0254] In an optional embodiment of the present invention, the historical feature set includes multiple first historical feature data arranged in chronological order, each first historical feature data corresponds to a time point; an extraction submodule is used to determine a first time point in the historical feature set where the first historical feature data is missing or the first historical feature data is an erroneous value; determine a second time point that records normal first historical feature data and is closest to the first time point; use the first historical feature data of the second time point to fill the first time point; divide the filled historical feature set according to a preset time interval to obtain at least one set of feature subsets; if all the first historical feature data in a feature subset are normal feature data, then the feature subset is used as a normal feature set; if the last first historical feature data in a feature subset is abnormal feature data, then the feature subset is used as an abnormal feature set.

[0255] In an optional embodiment of the present invention, the device further comprises:

[0256] The training set adding module is used to output the performance prediction result; receive the judgment result input for the performance prediction result; when the judgment result is correct, use the normal feature set and the abnormal feature set as the training feature set.

[0257] In an embodiment of the present invention, a historical feature set of a target optical module whose performance is to be predicted is first obtained; the historical feature set includes at least one first historical feature data recorded for the target optical module within a preset historical time period; the historical feature set is input into a preset prediction model to obtain a target prediction feature set; the target order parameter in the preset prediction model is determined by iterative calculation using a particle swarm optimization algorithm PSO to a preset number of iterations; and a performance prediction result for the target optical module is generated based on the target prediction feature set. Through the embodiment of the present invention, a prediction model for optical module performance is established by setting the number of iterations, which is suitable for automated scenarios and reduces the time and resources required to determine the order parameters in the prediction model, thereby improving the efficiency of predicting optical module performance.

[0258] An embodiment of the present invention further provides an electronic device, including a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program implements the above optical module performance prediction method when executed by the processor.

[0259] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above optical module performance prediction method is implemented.

[0260] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0261] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0262] Those skilled in the art will appreciate that the embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the embodiments of the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the embodiments of the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0263] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0264] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0265] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0266] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0267] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0268] The above is a detailed introduction to a method, device, electronic device and storage medium for predicting the performance of an optical module. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for predicting optical module performance, characterized in that: The method comprises: Acquire a historical feature set of a target optical module whose performance is to be predicted; the historical feature set includes at least one first historical feature data recorded for the target optical module within a preset historical time period; The historical feature set is input into a preset prediction model to obtain a target prediction feature set; the target order parameter in the preset prediction model is determined by iteratively calculating the particle swarm optimization algorithm PSO to a preset number of iterations; Generating a performance prediction result for the target optical module according to the target prediction feature set; Wherein, the method further comprises: Acquire at least one set of training feature sets, wherein the at least one set of training feature sets includes at least one second historical feature data recorded for at least one optical module, wherein the at least one optical module includes at least the target optical module; Obtaining a constraint range set for the order parameter, and searching the at least one set of training feature sets for a first order parameter corresponding to a minimum information criterion AIC value within the constraint range using an exhaustive algorithm; Using PSO to search for a second order parameter within the constraint range for the at least one set of training feature sets, and determining a corresponding number of iterations; the AIC value of the second order parameter is equal to the AIC value corresponding to the first order parameter; The number of iterations with the largest value among the numbers of iterations corresponding to all training feature sets is taken as the preset number of iterations; The PSO is used to iteratively calculate the order parameters in the preset prediction model, and when the number of iterations reaches the preset number of iterations, the target order parameters are output.

2. The method according to claim 1, characterized in that The step of inputting the historical feature set into a preset prediction model to obtain a target prediction feature set includes: Extracting a normal feature set and an abnormal feature set from the historical feature set; the first historical feature data in the normal feature set are all normal feature data, and the abnormal feature set includes at least one abnormal first historical feature data; Inputting the normal feature set into a preset prediction model to obtain a first prediction feature set; Inputting the abnormal feature set into a preset prediction model to obtain a second prediction feature set; Generating a performance prediction result for the target optical module according to the target prediction feature set includes: Acquire current characteristic data of the target optical module; A performance prediction result for the target optical module is generated according to the current feature data, the first prediction feature set, and the second prediction feature set.

3. The method according to claim 2, characterized in that The first prediction feature set and the second prediction feature set respectively include at least one prediction feature data, and generating a performance prediction result for the target optical module according to the current feature data, the first prediction feature set and the second prediction feature set includes: Determining a first standard deviation and a first mean value based on at least one predicted feature data in the first predicted feature set; and determining a second standard deviation and a second mean value based on at least one predicted feature data in the second predicted feature set; Acquire current characteristic data of the target optical module, and determine a first absolute value of a difference between the current characteristic data and the first average value, and a first ratio of the first absolute value of the difference to the first standard deviation; Determine a second absolute value of a difference between the abnormal feature data in the abnormal feature set and the second average value, and a second ratio of the second absolute value of the difference to the second standard deviation; When the first ratio is greater than the second ratio, generating a performance prediction result indicating an abnormality; When the first ratio is not greater than the second ratio, a performance prediction result indicating normality is generated.

4. The method according to claim 2, characterized in that: The historical feature set includes a plurality of first historical feature data arranged in chronological order, each first historical feature data corresponds to a time point; extracting a normal feature set and an abnormal feature set from the historical feature set includes: Determine a first time point in the historical feature set at which the first historical feature data is missing or has an erroneous value; Determine a second time point that has normal first historical feature data recorded and is closest to the first time point; Using the first historical feature data of the second time point to fill the first time point; According to a preset time interval, the filled historical feature set is segmented to obtain at least one set of feature subsets; If all the first historical feature data in a feature subset are normal feature data, the feature subset is used as a normal feature set; If the last first historical feature data in a feature subset is abnormal feature data, the feature subset is used as an abnormal feature set.

5. The method according to any one of claims 2 to 4, characterized in that: The method further comprises: Outputting the performance prediction result; receiving a determination result inputted for the performance prediction result; When the determination result is correct, the normal feature set and the abnormal feature set are used as training feature sets.

6. A device for predicting optical module performance, characterized in that: The device comprises: An acquisition module, for acquiring a historical feature set of a target optical module whose performance is to be predicted; the historical feature set includes at least one first historical feature data recorded for the target optical module within a preset historical time period; The prediction module inputs the historical feature set into a preset prediction model to obtain a target prediction feature set; the target order parameter in the preset prediction model is determined by iteratively calculating the particle swarm optimization algorithm PSO to a preset number of iterations; A performance prediction result generating module, used to generate a performance prediction result for the target optical module according to the target prediction feature set; A training module is used to obtain at least one set of training feature sets, wherein the at least one set of training feature sets includes at least one second historical feature data recorded for at least one optical module, and the at least one optical module at least includes the target optical module; obtain a constraint range set for the order parameter, and use an exhaustive algorithm to search for a first order parameter corresponding to a minimum information criterion AIC value within the constraint range for the at least one set of training feature sets; use PSO to search for a second order parameter within the constraint range for the at least one set of training feature sets, and determine the corresponding number of iterations; the AIC value of the second order parameter is equal to the AIC value corresponding to the first order parameter; among the numbers of iterations corresponding to all training feature sets, the number of iterations with the largest value is used as a preset number of iterations; use PSO to iteratively calculate the order parameters in the preset prediction model, and output the target order parameters when the number of iterations reaches the preset number of iterations.

7. An electronic device, characterized in that: The invention comprises a processor, a memory and a computer program stored in the memory and capable of running on the processor, wherein when the computer program is executed by the processor, the method for predicting the performance of an optical module as claimed in any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for predicting the performance of an optical module according to any one of claims 1 to 5 is implemented.

Images