WO2021248690A1 - 光通道性能劣化智能预警方法、装置、设备及存储介质 - Google Patents

光通道性能劣化智能预警方法、装置、设备及存储介质 Download PDF

Info

Publication number
WO2021248690A1
WO2021248690A1 PCT/CN2020/110176 CN2020110176W WO2021248690A1 WO 2021248690 A1 WO2021248690 A1 WO 2021248690A1 CN 2020110176 W CN2020110176 W CN 2020110176W WO 2021248690 A1 WO2021248690 A1 WO 2021248690A1
Authority
WO
WIPO (PCT)
Prior art keywords
performance
optical channel
monitored
model
prediction
Prior art date
Application number
PCT/CN2020/110176
Other languages
English (en)
French (fr)
Inventor
毕千筠
Original Assignee
烽火通信科技股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 烽火通信科技股份有限公司 filed Critical 烽火通信科技股份有限公司
Priority to BR112022019407A priority Critical patent/BR112022019407A2/pt
Publication of WO2021248690A1 publication Critical patent/WO2021248690A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B10/00—Transmission systems employing electromagnetic waves other than radio-waves, e.g. infrared, visible or ultraviolet light, or employing corpuscular radiation, e.g. quantum communication
    • H04B10/07—Arrangements for monitoring or testing transmission systems; Arrangements for fault measurement of transmission systems
    • H04B10/075—Arrangements for monitoring or testing transmission systems; Arrangements for fault measurement of transmission systems using an in-service signal
    • H04B10/077—Arrangements for monitoring or testing transmission systems; Arrangements for fault measurement of transmission systems using an in-service signal using a supervisory or additional signal
    • H04B10/0775—Performance monitoring and measurement of transmission parameters
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B10/00—Transmission systems employing electromagnetic waves other than radio-waves, e.g. infrared, visible or ultraviolet light, or employing corpuscular radiation, e.g. quantum communication
    • H04B10/07—Arrangements for monitoring or testing transmission systems; Arrangements for fault measurement of transmission systems
    • H04B10/075—Arrangements for monitoring or testing transmission systems; Arrangements for fault measurement of transmission systems using an in-service signal
    • H04B10/079—Arrangements for monitoring or testing transmission systems; Arrangements for fault measurement of transmission systems using an in-service signal using measurements of the data signal
    • H04B10/0795—Performance monitoring; Measurement of transmission parameters
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H04B17/373—Predicting channel quality or other radio frequency [RF] parameters
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H04B17/391—Modelling the propagation channel
    • H04B17/3913—Predictive models, e.g. based on neural network models

Definitions

  • the present invention relates to the field of communication technology, and in particular to an intelligent early warning method, device, equipment and storage medium for optical channel performance degradation.
  • the degradation of optical channel performance is an important factor that affects the quality of service transmission.
  • about 60% of transmission failures are caused by the performance of the optical channel; traditional network management monitors the transmission performance of the optical channel. It is achieved by manually setting some static thresholds for performance monitoring. When the optical channel-related performance value exceeds the threshold, the network manager will report an alarm; this method has two main drawbacks.
  • One is that the uniformly set static thresholds lack pertinence. The performance of the channel varies during different periods. The threshold setting is too high and it is not for monitoring purposes. If the threshold is set too low, some invalid alarms are frequently generated; the second is the lack of predictive ability of the performance degradation of the optical channel, and the performance and the performance before the failure can not be perceived in advance. The abnormal trend is a passive operation and maintenance afterwards.
  • the main purpose of the present invention is to provide an intelligent early warning method, device, equipment and storage medium for optical channel performance degradation, which aims to solve the inability to specifically determine the optical channel performance degradation in the prior art and lack of prediction of optical channel performance degradation. Technical issues of ability.
  • the present invention provides an intelligent early warning method for optical channel performance degradation.
  • the intelligent early warning method for optical channel performance degradation includes the following steps:
  • the performance trend prediction result it is determined whether the optical channel to be monitored has performance degradation, and when performance degradation occurs, an early warning process is performed.
  • the acquiring the performance data of the optical channel to be monitored in the telecommunications transmission network, and constructing the KPI performance prediction model of the key performance indicator of the optical channel to be monitored according to the performance data includes:
  • each performance prediction target and each target model algorithm construct the KPI performance prediction model of the key performance indicators of each optical channel to be monitored.
  • the construction of the key performance indicator KPI performance prediction model of each optical channel to be monitored according to each performance prediction target and each target model algorithm includes:
  • each performance prediction target and each target model algorithm construct the KPI performance prediction candidate baseline model of each optical channel to be monitored
  • Group training is performed on each optical channel group based on the optimal baseline model, and a KPI performance prediction model of each optical channel to be monitored is obtained.
  • the acquiring historical performance data of each optical channel to be monitored, and selecting an optimal baseline model from each KPI performance prediction candidate baseline model according to the historical performance data includes:
  • the KPI performance prediction candidate baseline model with the highest training accuracy is selected from each KPI performance prediction candidate baseline model as the optimal baseline model.
  • the performing periodic trend prediction on the performance of the optical channel to be monitored according to the KPI performance prediction model to obtain the performance trend prediction result includes:
  • the target performance data is input into the KPI performance prediction model to obtain the performance trend prediction result of the required period.
  • the judging whether the optical channel to be monitored has performance degradation according to the performance trend prediction result, and performing early warning processing when the performance degradation occurs includes:
  • the obtaining the time-sharing dynamic threshold for each time period from the performance trend prediction result includes:
  • the time-sharing dynamic threshold value of each time period is obtained according to the preset adjustment coefficient, the maximum prediction value and the minimum prediction value.
  • the following formulas are used to obtain the time-sharing dynamic threshold for each time period:
  • G t is the time-sharing dynamic threshold for each period
  • G max is the largest predicted value among all the predicted values for the current period
  • G min is the smallest predicted value among all predicted values for the current period
  • K is a preset adjustment coefficient
  • the comparing the performance trend prediction result with the time-sharing dynamic threshold to determine whether the current performance abnormality of the optical channel to be monitored occurs includes:
  • the performing early warning processing when the performance of the optical channel to be monitored is degraded includes:
  • the corresponding target early warning processing strategy is determined from the preset early warning strategy according to the number of abnormalities, and the early warning notification is performed according to the target early warning processing strategy.
  • the intelligent early warning method for optical channel performance degradation further includes:
  • the present invention also provides an intelligent early warning device for optical channel performance degradation
  • the optical channel performance degradation intelligent early warning device includes:
  • a model building module is used to obtain performance data of the optical channel to be monitored in the telecommunications transmission network, and construct a KPI performance prediction model for the key performance index of the optical channel to be monitored according to the performance data;
  • the prediction module is used to perform periodic trend prediction on the performance of the optical channel to be monitored according to the KPI performance prediction model, and obtain a performance trend prediction result;
  • the early warning module is used to determine whether the optical channel to be monitored has performance degradation according to the performance trend prediction result, and perform early warning processing when the performance degradation occurs.
  • model building module includes:
  • the performance acquisition module is used to acquire the performance data of a number of optical channels to be monitored in the telecommunication transmission network
  • Algorithm selection module used to determine each performance prediction target of each optical channel to be monitored according to various performance data, and select several corresponding target model algorithms from a preset database according to each performance prediction target;
  • the model generation module is used to construct the KPI performance prediction model of each optical channel to be monitored according to each performance prediction target and each target model algorithm.
  • model generation module includes:
  • the selection module is used to obtain historical performance data of each optical channel to be monitored, and select the optimal baseline model from the candidate baseline models for each KPI performance prediction according to the historical performance data;
  • the parameter acquisition module is used to acquire the channel parameters of a number of optical channels to be monitored
  • the dividing module is used to divide the optical channels to be monitored into optical channel groups with the same route and different wavelengths according to the channel parameters;
  • the group training module is used to perform group training on each optical channel group based on the optimal baseline model to obtain the KPI performance prediction model of each optical channel to be monitored.
  • the present invention also proposes an optical channel performance degradation intelligent early warning device.
  • the optical channel performance degradation intelligent early warning device includes: a memory, a processor, and a memory that is stored on the memory and can run on the processor.
  • the optical channel performance degradation intelligent early warning program is configured to implement the steps of the optical channel performance degradation intelligent early warning method as described above in the claims.
  • the present invention also provides a storage medium that stores an optical channel performance degradation intelligent warning program, and the optical channel performance degradation intelligent warning program is executed by a processor to realize the optical channel as described above. The steps of an intelligent early warning method for performance degradation.
  • the intelligent early warning method for optical channel performance degradation proposed in the present invention obtains performance data of the optical channel to be monitored in the telecommunication transmission network, and constructs the KPI performance prediction model of the key performance index of the optical channel to be monitored according to the performance data; according to the KPI performance
  • the prediction model performs periodic trend prediction on the performance of the optical channel to be monitored, and obtains the performance trend prediction result; judges whether the optical channel to be monitored has performance degradation according to the performance trend prediction result, and performs early warning processing when performance degradation occurs.
  • Targeted modeling of KPI performance indicators for optical channels in the transmission network ensures that the model always maintains good prediction accuracy in long-term engineering applications; at the same time, the long-term and short-term performance and trend prediction and analysis of optical channel KPI performance indicators are based on the model. Realize the early perception and automatic early warning of the performance degradation of the optical channel, and actively remind the operation and maintenance personnel to manually intervene in the degraded optical channel in advance to prevent problems before they occur.
  • FIG. 1 is a schematic diagram of a device structure of a hardware operating environment involved in a solution of an embodiment of the present invention
  • FIG. 2 is a schematic flowchart of a first embodiment of an intelligent early warning method for optical channel performance degradation according to the present invention
  • FIG. 4 is a schematic flowchart of a third embodiment of an intelligent early warning method for optical channel performance degradation according to the present invention.
  • FIG. 5 is a schematic flowchart of a fourth embodiment of an intelligent early warning method for optical channel performance degradation according to the present invention.
  • FIG. 7 is a schematic flowchart of a sixth embodiment of an intelligent early warning method for optical channel performance degradation according to the present invention.
  • FIG. 8 is a schematic flowchart of a seventh embodiment of an intelligent early warning method for optical channel performance degradation according to the present invention.
  • FIG. 9 is a schematic flowchart of an eighth embodiment of an intelligent early warning method for optical channel performance degradation according to the present invention.
  • FIG. 10 is a schematic flowchart of a ninth embodiment of an intelligent early warning method for optical channel performance degradation according to the present invention.
  • FIG. 11 is a schematic flowchart of a tenth embodiment of an intelligent early warning method for optical channel performance degradation according to the present invention.
  • Fig. 12 is a functional module diagram of the first embodiment of the intelligent early warning device for optical channel performance degradation according to the present invention.
  • the solution of the embodiment of the present invention is mainly: by obtaining the performance data of the optical channel to be monitored in the telecommunication transmission network, constructing the KPI performance prediction model of the key performance index of the optical channel to be monitored according to the performance data; and predicting the model according to the KPI performance Perform periodic trend prediction on the performance of the optical channel to be monitored to obtain a performance trend prediction result; determine whether the optical channel to be monitored has performance degradation according to the performance trend prediction result, and perform early warning processing when performance degradation occurs, which can be used to transmit Perform targeted modeling of KPI performance indicators of optical channels to ensure that the model always maintains good prediction accuracy in long-term engineering applications; at the same time, perform long-term and short-term performance and trend prediction and analysis on optical channel KPI performance indicators based on the model to achieve Early perception and automatic early warning of optical channel performance degradation, actively reminding operation and maintenance personnel to manually intervene in the degraded optical channel in advance, prevent problems before they occur, and solve the problem that the prior art cannot specifically determine the optical channel performance degradation and lack of light
  • the device may include: a processor 1001, such as a CPU, a communication bus 1002, a user-end interface 1003, a network interface 1004, and a memory 1005.
  • the communication bus 1002 is used to implement connection and communication between these components.
  • the user-side interface 1003 may include a display screen (Display) and an input unit such as a keyboard (Keyboard), and the optional user-side interface 1003 may also include a standard wired interface and a wireless interface.
  • the network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a Wi-Fi interface).
  • the memory 1005 may be a high-speed RAM memory, or a stable memory (Non-Volatile Memory), such as a disk memory.
  • the memory 1005 may also be a storage device independent of the aforementioned processor 1001.
  • FIG. 1 does not constitute a limitation on the device, and may include more or fewer components than those shown in the figure, or a combination of certain components, or different component arrangements.
  • the device of the present invention calls the optical channel performance degradation intelligent early warning program stored in the memory 1005 through the processor 1001, and executes the optical channel performance degradation intelligent early warning method provided by the embodiment of the present invention.
  • FIG. 2 is a schematic flowchart of a first embodiment of an intelligent early warning method for optical channel performance degradation according to the present invention.
  • the intelligent early warning method for optical channel performance degradation includes the following steps:
  • Step S10 Obtain performance data of the optical channel to be monitored in the telecommunications transmission network, and construct a KPI performance prediction model for the key performance index of the optical channel to be monitored according to the performance data.
  • the optical channel to be monitored is the current optical channel that needs to be monitored in the telecommunications transmission network, and the optical channel to be monitored may be one or multiple, which is not limited in this embodiment;
  • the performance data is data that can indicate the current optical channel performance of the optical channel to be monitored.
  • the performance data may be to obtain performance parameters over a period of time as performance data, for example, to obtain the original 15-minute granularity of the optical channel to be monitored. Performance data.
  • the performance data of the optical channel to be monitored can also be collected in other ways, which is not limited in this embodiment;
  • the performance parameter can be the error data of the optical channel, or the optical power data, or It can be optical signal-to-noise ratio (Optical Signal Noise Ratio, OSNR) data, of course, it can also be other types of data, or a combination of multiple data, which is not limited in this embodiment;
  • OSNR optical Signal Noise Ratio
  • KPI key performance indicator
  • the performance trend of the optical channel to be monitored in a certain period of time in the future can be obtained through the KPI performance prediction model, and the periodic trend prediction may be a long-term performance prediction. , It can also be a short-period performance prediction, which is specifically determined according to actual use requirements.
  • the performance cycle trend prediction is performed, the corresponding performance trend prediction result can be obtained, that is, the performance change trend in a certain period of time in the future.
  • Step S30 Determine whether the optical channel to be monitored has performance degradation according to the performance trend prediction result, and perform early warning processing when performance degradation occurs.
  • the performance trend prediction result can be analyzed to determine whether the performance trend has performance trend degradation, and if the degradation occurs, corresponding early warning processing is performed.
  • the early warning processing is Automatic early warning, that is, adopting different early warning notification strategies according to the degree of performance trend degradation.
  • the performance data of the optical channel to be monitored in the telecommunication transmission network is obtained, and the KPI performance prediction model of the key performance indicator of the optical channel to be monitored is constructed according to the performance data; Perform periodic trend prediction on the performance of the optical channel to be monitored to obtain the performance trend prediction result; determine whether the optical channel to be monitored has performance degradation according to the performance trend prediction result, and perform early warning processing when performance degradation occurs, which can be used for the transmission network optical channel Perform targeted modeling of KPI performance indicators to ensure that the model always maintains good prediction accuracy in long-term engineering applications; at the same time, perform long-term and short-term performance and trend prediction and analysis of optical channel KPI performance indicators based on the model to achieve optical channel performance Early detection of deterioration and automated early warning, proactively remind operation and maintenance personnel to manually intervene in the degraded optical channel in advance to prevent problems before they occur.
  • FIG. 3 is a schematic flowchart of the second embodiment of the optical channel performance degradation intelligent warning method according to the present invention.
  • the second embodiment of the optical channel performance degradation intelligent warning method according to the present invention is proposed based on the first embodiment.
  • the step S10 specifically includes the following steps:
  • Step S11 Obtain performance data of a number of optical channels to be monitored in the telecommunication transmission network.
  • Step S12 Determine each performance prediction target of each optical channel to be monitored according to each performance data, and select several corresponding target model algorithms from a preset database according to each performance prediction target.
  • the performance prediction target of each optical channel to be monitored is determined according to the performance data.
  • the performance prediction target corresponding to the error performance data is the channel data stream
  • the performance prediction target corresponding to the received and luminous power data is the optical signal output
  • the performance prediction target corresponding to the receiving frequency and OSNR data is the optical signal power and noise power within 0.1nm of the optical effective bandwidth, which is not limited in this embodiment
  • different performance prediction targets correspond to different model algorithms, which can be Pre-stored in a preset database, the model algorithm can choose a machine learning model based on autoregressive prediction, such as random forest model, ridge regression model, gradient boosting regression tree (Gradient Boosting Regression Trees, GBRT) model, and support vector regression model (Support Vector Regression, SVR), etc.
  • It can also be a deep learning model based on timing prediction, such as Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU), etc.
  • LSTM Long Short-Term Memory
  • GRU Gated Re
  • Step S13 Construct a KPI performance prediction model for each optical channel to be monitored according to each performance prediction target and each target model algorithm.
  • the performance prediction model can be based on the performance prediction target and the KPI performance prediction model corresponding to each target model algorithm component.
  • the performance data of a number of optical channels to be monitored in the telecommunications transmission network is obtained; the performance prediction targets of each optical channel to be monitored are determined according to the performance data, and the corresponding performance prediction targets are selected from the preset database according to the performance data.
  • the key performance indicator KPI performance prediction model of each optical channel to be monitored can be constructed according to the performance prediction target and the target model algorithm. The prediction accuracy of the model is valid for a long time.
  • FIG. 4 is a schematic flowchart of the third embodiment of the optical channel performance degradation intelligent warning method of the present invention.
  • the third embodiment of the optical channel performance degradation intelligent warning method of the present invention is proposed based on the second embodiment.
  • the step S13 specifically includes the following steps:
  • the prediction candidate baseline model is a model for predicting the KPI performance memory of the key performance indicator KPI of the optical channel to be monitored, which is constructed for different prediction targets and different model algorithm combinations. Different models have different technologies. As a result, when calculating the same prediction target, different algorithms will correspond to different prediction accuracy.
  • Step S132 Obtain historical performance data of each optical channel to be monitored, and select an optimal baseline model from each KPI performance prediction candidate baseline model according to the historical performance data.
  • the historical performance data is a record of the performance data of each optical channel to be monitored before the current time. It can be all previous performance data or historical performance data for a certain period of time before, for example, the past 30 days.
  • This embodiment does not impose any limitation on the historical data of the optical channel; the performance prediction candidate baseline model of each KPI is trained through the historical performance data to obtain the optimal baseline model. Generally, the model with the highest prediction accuracy is selected as the most accurate baseline model.
  • the optimal baseline model can also determine the optimal baseline model according to other methods, for example, the baseline model with the fastest prediction speed is used as the optimal baseline model, which is not limited in this embodiment.
  • Step S133 Obtain channel parameters of a number of optical channels to be monitored.
  • the channel parameters of the optical channel to be monitored are the basic data of the optical channel to be monitored, which can generally be obtained through the northbound interface provided by the transport network management system, or through a private protocol, or of course
  • the channel data is obtained in other ways, and this embodiment does not impose restrictions on this;
  • the channel data includes but is not limited to: optical channel name, speed, wavelength, source and sink network elements, and intermediate routing; it can also include the network management as an optical channel
  • the static threshold pre-configured for each performance index; of course, it may also include other data, which is not limited in this embodiment.
  • Step S134 According to various channel parameters, each optical channel to be monitored is divided into optical channel groups with the same route and different wavelengths.
  • the optical channels to be monitored can be grouped by each channel parameter, that is, the routes of the optical channels to be monitored are compared, and the optical channels with the same route and different wavelengths are divided into an optical channel group. The grouping is helpful for subsequent targeting. Group training is carried out to improve the accuracy and speed of the KPI performance prediction of the optical channel.
  • Step S135 Perform group training on each optical channel group based on the optimal baseline model to obtain a KPI performance prediction model for each optical channel to be monitored.
  • group training is performed on each optical channel group based on the optimal baseline model, that is, the corresponding optimal baseline model is selected for each group of optical channels for group prediction model training.
  • the optical channels In actual operation, generally Obtain the latest data of the KPI performance of this group of optical channels from the channel historical performance database as this group of training data. Since the optical channels of the same route generally have similar performance degradation rules, the optical channels can be generalized by grouping the models. On the one hand, it reduces the total amount of models and model training calculations required for optical channel performance prediction, and on the other hand, it can continue to maintain high model prediction accuracy at the same time.
  • the grouped generalization model trained by group training can be used as the KPI performance prediction model of each optical channel to be monitored, that is, the trained optical channel
  • the grouped generalization model of channel KPI performance prediction is used as the official model of the group of optical channel groups to be used online in the near future, and the model and its model-related information are stored in the optical channel model library of the device; the KPI performance prediction
  • the basic model information of the model includes but is not limited to: model ID, model classification, model usage, model algorithm, last training date, training accuracy, relevant baseline model version, and optical channel ID information for the model. Of course, it can also include more or There is less information, and this embodiment does not limit it.
  • the subsequent actual prediction accuracy of the group generalization model for the trained optical channel KPI performance prediction can be monitored; if the follow-up monitoring finds that the prediction accuracy of the model is significantly reduced, for example, the accuracy rate is lower than 80 %; or when the routing or configuration of the related optical channel is changed; at this time, it is necessary to restart the retraining of the generalization model of the related optical channel KPI performance grouping to further improve the prediction accuracy of the model.
  • the KPI performance prediction candidate baseline model of the key performance index KPI of each optical channel to be monitored is constructed according to the performance prediction target and each target model algorithm; the historical performance data of each optical channel to be monitored is acquired, and the historical performance data
  • the performance data selects the optimal baseline model from the candidate baseline models of each KPI performance prediction; obtains the channel parameters of a number of optical channels to be monitored; divides the optical channels to be monitored into optical channel groups with the same route and different wavelengths according to the channel parameters;
  • the optimal baseline model performs group training on each optical channel group, and obtains the KPI performance prediction model of each optical channel to be monitored, which can reduce the total amount of model and model training calculation required for optical channel performance prediction, and further improve the prediction of the model. Accuracy.
  • FIG. 5 is a schematic flowchart of the fourth embodiment of the optical channel performance degradation intelligent warning method of the present invention.
  • the fourth embodiment of the optical channel performance degradation intelligent warning method of the present invention is proposed based on the third embodiment.
  • the step S132 specifically includes the following steps:
  • Step S1321 obtain historical performance data of each optical channel to be monitored.
  • the historical performance data is the performance data record generated by the optical channel to be monitored in the past period of time, or it may be all the performance data records of the optical channel to be monitored before the current moment, which is not included in this embodiment. Be restricted.
  • Step S1322 Train each KPI performance prediction candidate baseline model according to the historical performance data to obtain the training accuracy of each KPI performance prediction candidate baseline model.
  • each KPI performance prediction candidate baseline model is trained, and the historical performance data is substituted into each KPI performance prediction candidate baseline model, and the obtained predicted value is compared with the true value after calculation.
  • the performance of each KPI predicts the training accuracy of the candidate baseline model.
  • the candidate baseline model for KPI performance prediction can be a short-period prediction model and a long-period prediction model;
  • the short-period prediction model can predict the daily trend of the 15-minute granular performance of the optical channel KPI performance, and the model goal is from the past N days (N generally takes ⁇ 10 days) historical optical channel performance data (15-minute granularity) predicts all 15-minute granular performance values on day N+1;
  • the long-period prediction model can measure the M-day average daily trend of optical channel KPI performance For forecasting, M generally takes 3 or 5 days (same as above) in order to not only better reflect the trend of the optical channel, but also have better sensitivity to changes in the trend.
  • the long-term forecast model target is from the past N days ( (N generally takes ⁇ 30 days) optical channel historical data to predict the performance trend value of the M-day moving average of N+M days, that is, the average daily average value, average daily maximum value, and average daily minimum value within M days.
  • Step S1323 Select the KPI performance prediction candidate baseline model with the highest training accuracy from the KPI performance prediction candidate baseline models as the optimal baseline model.
  • each category model can be subdivided into multiple sub-category models; for the training of all candidate models, the historical performance data of optical channels of different projects can be selected for more than 3 months. Conduct training.
  • the training accuracy (ie accuracy rate) of the model is uniformly calculated as follows:
  • Model deviation rate
  • Model accuracy rate the number of optical channels with a deviation rate of less than 20%/the number of all predicted optical channels
  • each KPI performance prediction candidate baseline model is trained according to the historical performance data to obtain the training accuracy of each KPI performance prediction candidate baseline model;
  • the KPI performance prediction candidate baseline model with the highest training accuracy is selected as the optimal baseline model, which can ensure the accuracy of performance prediction and the long-term effectiveness of the model accuracy, providing a detection basis for subsequent optical channel performance anomaly detection .
  • FIG. 6 is a schematic flowchart of the fifth embodiment of the optical channel performance degradation intelligent warning method of the present invention.
  • the fifth embodiment of the optical channel performance degradation intelligent warning method of the present invention is proposed based on the first embodiment.
  • the step S20 specifically includes the following steps:
  • Step S21 Obtain the model category and model usage of the KPI performance prediction model.
  • the model category is a category classified according to the type of the KPI performance prediction model, and according to the model category, it can be determined whether the KPI performance prediction model is a short-period prediction model or a long-period prediction model;
  • the model usage is the usage divided according to the prediction objects of the KPI performance prediction model. According to the model usage, it can be determined whether the KPI performance prediction model is an error prediction model, a light emission power prediction model, or a signal-to-noise ratio prediction. Models, or other types of performance parameter prediction models, are not limited in this embodiment.
  • Step S22 Read the target performance data of the required period from the historical performance database of each optical channel to be monitored according to the model category and the purpose of the model.
  • matching performance data can be obtained from the historical performance database through the model category and the model usage, that is, the target performance data in the required period corresponding to the model category and the model usage .
  • the optical channel model library is generally queried.
  • the optical channel model library is a database for storing various optical channel models.
  • the channel model library can be stored locally or in the cloud.
  • This embodiment does not impose restrictions on this; obtain the optical channel grouping generalization model corresponding to the KPI performance prediction of the related optical channel, specifically through model classification and model Use and whether the optical channel ID applicable to the model matches the optical channel ID for model selection; according to the different model types and uses of the obtained optical channel grouping prediction model, read the optical channel historical performance database of the device
  • the original performance data of the past N days (N generally takes 10 days) for the preset time required by the model, generally 15 minutes of original performance data, that is, as the target performance data; of course, it can also be other preset times, this embodiment
  • the matching model is also determined through the optical channel model library.
  • the different model types and uses of the obtained optical channel grouping generalization model from The M-day aggregated data of the past N days (N generally takes 30 days) required to read the model in the historical performance database of the optical channel is used as the target performance data.
  • Step S23 Input the target performance data into the KPI performance prediction model to obtain the performance trend prediction result of the required period.
  • the output result obtained is the performance trend prediction result of the required period; for short-period performance prediction, dynamic loading and allowing all The KPI performance prediction model is described.
  • the output result of the model is obtained.
  • the result is the performance prediction value of the relevant optical channel at the preset time on the N+1 day; for long-term performance prediction, the model is obtained after the model runs.
  • the output result is the average daily trend prediction value of the relevant optical channel’s future M-day KPI performance; after the performance trend prediction result is obtained, the corresponding forecast date and result are generally saved in the optical channel prediction database. In order to facilitate subsequent further analysis and calculation.
  • the model category and model usage of the KPI performance prediction model are obtained; according to the model category and the model usage, the required period of target performance data is read from the historical performance database of each optical channel to be monitored ; Input the target performance data into the KPI performance prediction model to obtain the performance trend prediction result of the required period, which can perform targeted performance prediction according to actual demand, which improves the flexibility and accuracy of performance prediction .
  • FIG. 7 is a schematic flowchart of the sixth embodiment of the optical channel performance degradation intelligent warning method of the present invention. As shown in FIG. 7, the sixth embodiment of the optical channel performance degradation intelligent warning method of the present invention is proposed based on the first embodiment.
  • the step S30 specifically includes the following steps:
  • Step S31 Obtain the time-sharing dynamic threshold value of each time period from the performance trend prediction result.
  • time-sharing dynamic threshold is a preset threshold corresponding to the performance trend prediction value of each time period, and the time-sharing dynamic threshold value of each time period is generally different and dynamically changing.
  • Step S32 Compare the performance trend prediction result with the time-sharing dynamic threshold, and determine whether the optical channel to be monitored has a current performance abnormality.
  • the optical channel to be monitored is generally abnormal in real time Detection, thereby improving the efficiency of data processing and the timeliness of early warning of optical channel performance.
  • Step S33 Obtain a trend prediction value from the performance trend prediction result, and determine whether the optical channel to be monitored has an abnormal trend according to whether the trend prediction value enters a preset performance degradation risk zone.
  • the performance trend prediction result includes a trend performance prediction for a certain time period in the future
  • the preset performance degradation risk area is a preset dynamic risk area that assists performance trend degradation judgment
  • the trend prediction Whether the value enters the preset performance degradation risk zone can determine whether the optical channel to be monitored has an abnormal trend.
  • the trend prediction value is obtained from the performance trend prediction result, generally by obtaining the M-day pre-aggregated data of the relevant optical channel KPI performance index in the past N days from the historical performance database as the trend prediction value.
  • the performance degradation risk area can be divided into the online middle risk area, the online high risk area, the offline middle risk area and the offline high risk area; of course, it can also be set to other types of performance degradation risk areas.
  • the preset performance degradation risk Areas are generally calculated by calculating the trend prediction value to obtain the upper-line mid-risk area, the upper-line high-risk area, the down-line mid-risk area, and the offline high-risk area as examples:
  • R1 Average daily average value of actual KPI performance over the past N days
  • R2 the actual average daily maximum value of KPI performance in the past N days
  • R3 the actual average daily minimum value of KPI performance in the past N days
  • R4 Min(R1+(R2-R1) ⁇ 2, online static threshold)
  • R5 Max(R1-(R1-R3) ⁇ 2, offline static threshold)
  • the calculation of the actual average daily maximum value, average daily minimum value and average daily average value of actual KPI performance needs to eliminate abnormal points, and perform M-day trend prediction value detection on each KPI performance of the monitored optical channel to determine whether there is an abnormal trend;
  • the judgment of abnormal trend is divided into the following four categories:
  • Type A anomaly Determine whether the average daily maximum value P max _i of the future M-day optical channel KPI performance enters the upper risk zone, and whether the average daily average P avg _i of the future M-day optical channel KPI performance continues to rise.
  • Type B anomaly Determine whether the average daily minimum value P min _i of the future M-day optical channel KPI performance enters the lower limit of the risk zone, and whether the average daily average P avg _i of the future M-day optical channel KPI performance continues to decline.
  • Type C anomaly Determine whether the average daily maximum value P max _i of the future M-day optical channel KPI performance enters the upper limit high risk area.
  • Type D anomaly Determine whether the average daily maximum value P min _i of the future M-day optical channel KPI performance enters the lower limit high risk area.
  • the error rate trend is determined and tested for type A and C abnormalities, and the type A, B, C, and D abnormalities are determined and tested for the light-emitting power trend. , Perform judgment and detection of B and D abnormalities on OSNR trends;
  • the optical channel performance trend abnormal event is usually recorded and saved in the optical channel performance abnormal record;
  • the optical channel performance trend abnormality records the following information: related KPI indicators, abnormal classification (A/B/C/ Type D anomaly), detection date, related optical channel ID and name, and M-day trend prediction value and risk determination value used in abnormal determination; of course, it can also include more or less information, which is in this embodiment No restrictions.
  • Step S34 When the current performance abnormality and/or the trend abnormality occur, it is determined that the performance of the optical channel to be monitored has deteriorated.
  • Step S35 Perform an early warning process when the performance of the optical channel to be monitored is degraded.
  • the time-sharing dynamic threshold value of each period is obtained from the performance trend prediction result; the performance trend prediction result is compared with the time-sharing dynamic threshold value to determine whether the optical channel to be monitored appears
  • the current performance is abnormal; the trend prediction value is obtained from the performance trend prediction result, and it is determined whether the optical channel to be monitored has a trend abnormality according to whether the trend prediction value enters the preset performance degradation risk zone; when the current performance abnormality and When the trend is abnormal, it is determined that the performance of the optical channel to be monitored is degraded; when the performance of the optical channel to be monitored is degraded, early warning processing is performed, which can comprehensively detect the abnormality in the optical channel to be monitored, thereby improving
  • the performance detection accuracy of the optical channel to be monitored realizes the early perception and automatic early warning of the performance degradation of the optical channel.
  • FIG. 8 is a schematic flowchart of the seventh embodiment of the optical channel performance degradation intelligent warning method of the present invention.
  • the seventh embodiment of the optical channel performance degradation intelligent warning method of the present invention is proposed based on the sixth embodiment.
  • the step S31 specifically includes the following steps:
  • Step S311 Obtain all predicted values of the current period in each period from the performance trend prediction result.
  • the performance trend prediction result corresponds to the performance prediction value of each time period, and the current time period prediction value is the predicted value of each time period in the preset time period.
  • Step S312 Determine the maximum prediction value and the minimum prediction value from all the prediction values of the current period.
  • Zhang Gong can determine the largest predicted value and the smallest predicted value from all the predicted values of the current period.
  • Step S313 Obtain a time-sharing dynamic threshold for each time period according to the preset adjustment coefficient, the maximum predicted value and the minimum predicted value.
  • the preset adjustment coefficient is a preset adjustment coefficient for calculating the time-sharing dynamic threshold, and the preset adjustment coefficient, the maximum predicted value, and the minimum predicted value can be calculated to obtain the score for each time period. Time dynamic threshold.
  • step S313 specifically includes the following steps:
  • the preset adjustment coefficient the maximum predicted value and the minimum predicted value, the following formulas are used to obtain the time-sharing dynamic thresholds of each time period:
  • G t is the time-sharing dynamic threshold for each period
  • G max is the largest predicted value among all the predicted values for the current period
  • G min is the smallest predicted value among all predicted values for the current period
  • K is a preset adjustment coefficient
  • the time-sharing dynamic threshold of the N+1 day of the optical channel can be calculated based on the predicted value.
  • the time-sharing The time period is defined as a time period every H hours. H is generally 1/2/3/4 hours, and the default value is 2 hours.
  • the adjustment coefficient K can be a fixed value between 1.0 and 2.0, and the initial value can be 1.0.
  • Subsequent adjustments can be made according to the actual monitoring effect of the optical channel; of course it can also be other values, which are not limited in this embodiment; after obtaining the time-sharing dynamic threshold for each period, it can be stored to generate the optical channel performance
  • the time-sharing dynamic threshold table which can be used in the subsequent analysis of the current abnormal performance of the optical channel.
  • all the prediction values of the current period of each period are obtained from the performance trend prediction results; the maximum prediction value and the minimum prediction value are determined from all the prediction values of the current period; according to the preset adjustment coefficient,
  • the maximum prediction value and the minimum prediction value obtain the time-sharing dynamic threshold values of each time period, and the dynamic threshold value can be set in time periods, which further improves the accuracy of performance prediction and serves as a basis for judging the performance degradation of the optical channel.
  • FIG. 9 is a schematic flowchart of the eighth embodiment of the intelligent early warning method for optical channel performance degradation according to the present invention.
  • the eighth embodiment of the intelligent early warning method for optical channel performance degradation according to the present invention is proposed based on the sixth embodiment.
  • the step S32 specifically includes the following steps:
  • Step S321 Obtain current performance data corresponding to the current moment from the performance trend prediction result, and obtain the current dynamic threshold value at the current moment from the time-sharing dynamic threshold value.
  • the performance trend prediction result stores performance trend prediction data for each time period, from which performance data matching the current moment can be obtained, and the current moment corresponding to the time-sharing dynamic threshold can be obtained from the time-sharing dynamic threshold.
  • the current dynamic threshold stores performance trend prediction data for each time period, from which performance data matching the current moment can be obtained, and the current moment corresponding to the time-sharing dynamic threshold can be obtained from the time-sharing dynamic threshold.
  • Step S322 Determine whether the current performance data is invalid data.
  • the invalid data generally includes missing performance values, too large or too small values, illegal zero values or invalid Null values, etc. Of course, it can also include other types of invalid data, which is not limited in this embodiment;
  • the KPI performance data of the current 15-minute granularity of the monitored optical channel is generally detected as current performance data to determine whether the current performance data is invalid data, and if it is invalid data, it is discarded without any processing.
  • Step S323 When the current data is not invalid data, obtain the optical channel performance static threshold of the optical channel to be monitored.
  • the static threshold of optical channel performance is a fixed threshold that is preset for optical channel performance; generally, from the optical channel KPI performance time-sharing dynamic threshold cache table, read the optical channel corresponding to the current moment Current dynamic threshold, and read the static threshold of optical channel performance obtained from the network management from the basic optical performance data.
  • Step S324 Compare the current performance data with the optical channel performance static threshold and the current dynamic threshold respectively.
  • Step S325 When the current performance data is not greater than the optical channel performance static threshold value and the current dynamic threshold value, it is determined that the current performance abnormality of the optical channel to be monitored does not occur.
  • Step S326 When the current performance data is greater than the optical channel performance static threshold and/or the current dynamic threshold, it is determined that the current performance of the optical channel to be monitored is abnormal.
  • the current performance data is greater than one or two of the optical channel performance static threshold and the current dynamic threshold, it can be determined that the optical channel to be monitored has a current performance abnormality.
  • a current abnormal performance event of the optical channel can be recorded and saved in the abnormal record of the optical channel performance library; the current abnormal performance event of the optical channel records the following information: related KPI performance indicators, abnormalities Classification (static over-limit, dynamic over-limit), abnormal occurrence time, related optical channel ID and name, abnormal performance value, time-sharing performance dynamic threshold and static threshold used for abnormal determination.
  • related KPI performance indicators abnormalities Classification (static over-limit, dynamic over-limit), abnormal occurrence time, related optical channel ID and name, abnormal performance value, time-sharing performance dynamic threshold and static threshold used for abnormal determination.
  • the current performance data corresponding to the current moment is obtained from the performance trend prediction result, and the current dynamic threshold value of the current moment is obtained from the time-sharing dynamic threshold value; and the current performance data is determined Whether it is invalid data; when the current data is not invalid data, obtain the optical channel performance static threshold of the optical channel to be monitored; compare the current performance data with the optical channel performance static threshold and the current dynamic Threshold comparison; when the current performance data is not greater than the optical channel performance static threshold and the current dynamic threshold, it is determined that the optical channel to be monitored has no current performance abnormality; when the current performance data is greater than the optical channel
  • the channel performance static threshold value and/or the current dynamic threshold value are determined, the current performance abnormality of the optical channel to be monitored can be determined, which can more comprehensively determine the current performance abnormality, and further improve the accuracy of optical channel performance detection.
  • FIG. 10 is a schematic flowchart of the ninth embodiment of the optical channel performance degradation intelligent early warning method of the present invention.
  • the ninth embodiment of the optical channel performance degradation intelligent early warning method of the present invention is proposed based on the sixth embodiment.
  • the step S35 specifically includes the following steps:
  • Step S351 When the performance trend of the optical channel to be monitored is degraded, obtain the current abnormal performance and/or abnormal trend times.
  • the number of abnormalities is the total number of abnormalities of current performance abnormalities and abnormal trends when the optical channel to be monitored has a performance trend degradation. Of course, there may be only one abnormality, and the abnormal number of times is at this time. Is the total number of exceptions for this exception.
  • Step S352 Determine a corresponding target early warning processing strategy from a preset early warning strategy according to the number of abnormalities, and perform an early warning notification according to the target early warning processing strategy.
  • the preset early warning strategy may include two parts: one is the trigger strategy for when and what level of early warning of optical channel KPI performance abnormal events are triggered, and the other is the trigger for early warning of optical channel performance abnormalities.
  • the following notification strategy; the basic early warning trigger strategy is defined as follows:
  • the basic warning notification strategy includes: email notification, short message notification, system notification, and combined notification. Of course, it can also use other methods, such as telephone notification, sound and light alarm, etc.
  • This embodiment does not limit this; generally, when receiving After the current abnormality or trend of optical channel KPI performance is notified, the early warning monitoring work is automatically started; the recent abnormal records related to the currently reported abnormal optical channel are analyzed and counted from the abnormal event library, and triggered according to the pre-defined early warning in the previous step The strategy determines whether the recent accumulated abnormalities of the current optical channel meet the trigger conditions defined by the strategy, and for the optical channel performance and trend deterioration events that meet the early warning trigger conditions, automatic early warning notifications are performed according to the early warning notification strategy.
  • the corresponding target warning is determined from the preset warning strategy according to the abnormality times Processing strategy, according to the target early warning processing strategy to provide early warning notifications, to perform targeted early warning processing operations, to realize early perception and automatic early warning of optical channel performance degradation, and to actively remind operation and maintenance personnel to manually intervene in the degraded optical channel in advance. Take precautions before they happen.
  • FIG. 11 is a schematic flowchart of the tenth embodiment of the intelligent early warning method for optical channel performance degradation according to the present invention. As shown in FIG. 11, the tenth embodiment of the intelligent early warning method for optical channel performance degradation according to the present invention is proposed based on the second embodiment. In this embodiment, the step S11 specifically includes the following steps:
  • Step S111 Perform an invalid data check on the performance data.
  • invalid data can be sorted out, and the accuracy of performance detection can be further improved; the invalid data generally includes missing performance values, too large or too small values, and illegal zero values. Or invalid Null values, etc., of course, can also include other types of invalid data, which is not limited in this embodiment.
  • Step S112 When target invalid data appears in the performance data, obtain the effective performance value of the last preset time or the next preset time of the target invalid data.
  • the invalid data can be replaced by obtaining the effective performance value at the previous preset time or the next preset time of the target invalid data, thereby ensuring The accuracy of the forecast.
  • the date and time fields related to the original performance can also be standardized to unify the data accuracy, and convert it into a time data type that is easy to process in the development language, and save the processed optical channel 15-minute KPI performance data to In the optical channel historical performance database, the accuracy of performance detection is further improved, and the detection speed and efficiency are improved.
  • Step S113 Replace the target invalid data with the effective performance value, and use the replaced performance data as new performance data.
  • the 15-minute performance data of the optical channel can be pre-aggregated and calculated daily.
  • the pre-aggregation adopts a sliding time window. Based on the original 15-minute performance data, the KPI performance of each optical channel in the past M days is calculated. Average data; in order to not only reflect the trend of optical channel performance, but also maintain good sensitivity to trend changes, M generally takes 3 or 5 days, and the default is 5 days; data that needs to be pre-aggregated include: past M The average daily average value, average daily maximum value, and average daily minimum value within the day; it should be noted that some occasional abnormal points need to be eliminated during the pre-aggregation calculation.
  • Replacing the target invalid data with the effective performance value and using the replaced performance data as new performance data can further ensure the accuracy of performance prediction, ensure the long-term validity of the prediction accuracy of the model, and improve the detection speed and performance. efficient.
  • the present invention further provides an intelligent early warning device for optical channel performance degradation.
  • Fig. 12 is a functional module diagram of a first embodiment of an intelligent early warning device for optical channel performance degradation according to the present invention.
  • the optical channel performance degradation intelligent early warning device includes:
  • the model construction module 10 is configured to obtain performance data of the optical channel to be monitored in the telecommunication transmission network, and construct a KPI performance prediction model of the key performance index of the optical channel to be monitored according to the performance data.
  • the prediction module 20 is configured to perform periodic trend prediction on the performance of the optical channel to be monitored according to the KPI performance prediction model, and obtain a performance trend prediction result.
  • the early warning module 30 is configured to determine whether the optical channel to be monitored has performance degradation according to the performance trend prediction result, and perform early warning processing when the performance degradation occurs.
  • model construction module 10 includes:
  • the performance acquisition module is used to acquire the performance data of a number of optical channels to be monitored in the telecommunication transmission network
  • Algorithm selection module used to determine each performance prediction target of each optical channel to be monitored according to various performance data, and select several corresponding target model algorithms from a preset database according to each performance prediction target;
  • the model generation module is used to construct the KPI performance prediction model of each optical channel to be monitored according to each performance prediction target and each target model algorithm.
  • model generation module includes:
  • the baseline model building module is used to construct the KPI performance prediction candidate baseline model of each optical channel to be monitored according to each performance prediction target and each target model algorithm;
  • the selection module is used to obtain historical performance data of each optical channel to be monitored, and select the optimal baseline model from the candidate baseline models for each KPI performance prediction according to the historical performance data;
  • the parameter acquisition module is used to acquire the channel parameters of a number of optical channels to be monitored
  • the dividing module is used to divide the optical channels to be monitored into optical channel groups with the same route and different wavelengths according to the channel parameters;
  • the group training module is used to perform group training on each optical channel group based on the optimal baseline model to obtain the KPI performance prediction model of each optical channel to be monitored.
  • each functional module of the optical channel performance degradation intelligent early warning device can refer to the various embodiments of the optical channel performance degradation intelligent early warning method of the present invention, which will not be repeated here.
  • the embodiment of the present invention also proposes a storage medium
  • the storage medium may be a computer-readable non-volatile storage medium, of course, it may also be other types of storage media, which is not limited in this embodiment;
  • the storage medium stores an intelligent early warning program for optical channel performance degradation, and when the optical channel performance degradation intelligent early warning program is executed by a processor, the implementation method for the optical channel performance degradation intelligent early warning provided in the above embodiment of the present invention is implemented.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Electromagnetism (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Quality & Reliability (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Optical Communication System (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

本发明公开了一种光通道性能劣化智能预警方法、装置、设备及存储介质，所述方法通过获取电信传送网中待监测光通道的性能数据，根据性能数据构建待监测光通道的关键绩效指标KPI性能预测模型；根据KPI性能预测模型对待监测光通道的性能进行周期趋势预测，获得性能趋势预测结果；根据性能趋势预测结果判断待监测光通道是否出现性能趋势劣化，在出现性能趋势劣化时，进行预警处理，能够对传送网光通道KPI性能指标进行针对性建模，保障了模型在工程长期应用中始终保持较好的预测精度；同时基于模型对光通道KPI性能指标进行长短期性能与趋势的预测和分析，实现对光通道性能劣化的提前感知和自动化预警。

Description

光通道性能劣化智能预警方法、装置、设备及存储介质 技术领域
本发明涉及通信技术领域，尤其涉及一种光通道性能劣化智能预警方法、装置、设备及存储介质。
背景技术
在电信光传送网中，光通道性能劣化是影响业务传输质量的一个重要因素，据统计大约60％的传输故障是由光通道的性能原因引起的；传统网管对于光通道传输性能的监控，一般是通过手工设置一些性能监控的静态阈值来实现，当光通道相关性能值超过该阈值时网管就报告警；这种方式有两个主要缺陷，一是统一设置的静态阈值缺乏针对性，不同光通道不同时期其性能表现各异，阈值设置过高起不到监控目的，设置过低又频繁产生一些无效告警；二是缺乏对光通道性能劣化的预测能力，不能提前感知故障发生前的性能和趋势的异常，是一种事后被动式运维。
发明内容
本发明的主要目的在于提供一种光通道性能劣化智能预警方法、装置、设备及存储介质，旨在解决现有技术中不能针对性的确定光通道性能劣化，且缺乏对光通道性能劣化的预测能力的技术问题。
第一方面，本发明提供一种光通道性能劣化智能预警方法，所述光通道性能劣化智能预警方法包括以下步骤：
获取电信传送网中待监测光通道的性能数据，根据所述性能数据构建所述待监测光通道的关键绩效指标KPI性能预测模型；
根据KPI性能预测模型对所述待监测光通道的性能进行周期趋势预测，获得性能趋势预测结果；
根据性能趋势预测结果判断所述待监测光通道是否出现性能劣化，在出现性能劣化时，进行预警处理。
可选地，所述获取电信传送网中待监测光通道的性能数据，根据所述性 能数据构建所述待监测光通道的关键绩效指标KPI性能预测模型，包括：
获取电信传送网中若干待监测光通道的性能数据；
根据各性能数据确定各待监测光通道的各性能预测目标，根据各性能预测目标从预设数据库中选取对应的若干目标模型算法；
根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测模型。
可选地，所述根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测模型，包括：
根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测候选基线模型；
获取各待监测光通道的历史性能数据，根据所述历史性能数据从各KPI性能预测候选基线模型中选取最优基线模型；
获取若干待监测光通道的通道参数；
根据各通道参数将各待监测光通道划分为相同路由不同波长的光通道组；
基于所述最优基线模型对各光通道组进行分组训练，获得各待监测光通道的KPI性能预测模型。
可选地，所述获取各待监测光通道的历史性能数据，根据所述历史性能数据从各KPI性能预测候选基线模型中选取最优基线模型，包括：
获取各待监测光通道的历史性能数据；
根据所述历史性能数据对各KPI性能预测候选基线模型进行训练，获得各KPI性能预测候选基线模型的训练精度；
从各KPI性能预测候选基线模型中选取训练精度最高的KPI性能预测候选基线模型作为最优基线模型。
可选地，所述根据KPI性能预测模型对所述待监测光通道的性能进行周期趋势预测，获得性能趋势预测结果，包括：
获取KPI性能预测模型的模型类别和模型用途；
根据所述模型类别和所述模型用途从各待监测光通道的历史性能库中读取所需周期的目标性能数据；
将所述目标性能数据输入至所述KPI性能预测模型中，获得所述所需周 期的性能趋势预测结果。
可选地，所述根据性能趋势预测结果判断所述待监测光通道是否出现性能劣化，在出现性能劣化时，进行预警处理，包括：
从所述性能趋势预测结果中获得各个时段的分时动态阈值；
将所述性能趋势预测结果与所述分时动态阈值比较，判断所述待监测光通道是否出现当前性能异常；
从所述性能趋势预测结果中获得趋势预测值，根据所述趋势预测值是否进入了预设性能劣化风险区判断所述待监测光通道是否出现趋势异常；
在出现当前性能异常和/或趋势异常时，判定所述待监测光通道出现性能劣化；
在所述待监测光通道出现性能劣化时，进行预警处理。
可选地，所述从所述性能趋势预测结果中获得各个时段的分时动态阈值，包括：
从所述性能趋势预测结果中获得各个时段的所有本时段预测值；
从所述所有本时段预测值中确定最大预测值和最小预测值；
根据预设调整系数、所述最大预测值和所述最小预测值获得各个时段的分时动态阈值。
可选地，根据预设调整系数、所述最大预测值和所述最小预测值利用下述公式获得各个时段的分时动态阈值：
G t＝G max+|G max-G min|×K
其中，G t为各个时段的分时动态阈值，G max为所述所有本时段预测值中的最大预测值，G min为所述所有本时段预测值中的最小预测值，K为预设调整系数。
可选地，所述将所述性能趋势预测结果与所述分时动态阈值比较，判断所述待监测光通道是否出现当前性能异常，包括：
从所述性能趋势预测结果中获取当前时刻对应的当前性能数据，并从所述分时动态阈值中获取所述当前时刻的当前动态阈值；
判断所述当前性能数据是否为无效数据；
在所述当前数据不为无效数据时，获取所述待监测光通道的光通道性能静态阈值；
将所述当前性能数据分别与所述光通道性能静态阈值及所述当前动态阈值的比较；
在所述当前性能数据不大于所述光通道性能静态阈值及所述当前动态阈值时，判定所述待监测光通道未出现当前性能异常；
在所述当前性能数据大于所述光通道性能静态阈值和/或所述当前动态阈值时，判定所述待监测光通道出现当前性能异常。
可选地，所述在所述待监测光通道出现性能劣化时，进行预警处理，包括：
在所述待监测光通道出现性能趋势劣化时，获取当前性能异常和/或趋势异常的异常次数；
根据所述异常次数从预设预警策略中确定对应的目标预警处理策略，根据所述目标预警处理策略进行预警通知。
可选地，所述获取电信传送网中若干待监测光通道的性能数据之后，所述光通道性能劣化智能预警方法还包括：
对所述性能数据进行无效数据检查；
在所述性能数据中出现目标无效数据时，获取所述目标无效数据的上一预设时刻或下一预设时刻的有效性能值；
将所述目标无效数据替换为所述有效性能值，将替换后的性能数据作为新的性能数据。
第二方面，本发明还提出一种光通道性能劣化智能预警装置，所述光通道性能劣化智能预警装置包括：
模型构建模块，用于获取电信传送网中待监测光通道的性能数据，根据所述性能数据构建所述待监测光通道的关键绩效指标KPI性能预测模型；
预测模块，用于根据KPI性能预测模型对所述待监测光通道的性能进行周期趋势预测，获得性能趋势预测结果；
预警模块，用于根据性能趋势预测结果判断所述待监测光通道是否出现性能劣化，在出现性能劣化时，进行预警处理。
可选地，所述模型构建模块包括：
性能获取模块，用于获取电信传送网中若干待监测光通道的性能数据；
算法选取模块，用于根据各性能数据确定各待监测光通道的各性能预测目标，根据各性能预测目标从预设数据库中选取对应的若干目标模型算法；
模型生成模块，用于根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测模型。
可选地，所述模型生成模块包括：
基线模型构建模块，用于根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测候选基线模型；
选取模块，用于获取各待监测光通道的历史性能数据，根据所述历史性能数据从各KPI性能预测候选基线模型中选取最优基线模型；
参数获取模块，用于获取若干待监测光通道的通道参数；
划分模块，用于根据各通道参数将各待监测光通道划分为相同路由不同波长的光通道组；
分组训练模块，用于基于所述最优基线模型对各光通道组进行分组训练，获得各待监测光通道的KPI性能预测模型。
第三方面，本发明还提出一种光通道性能劣化智能预警设备，所述光通道性能劣化智能预警设备包括：存储器、处理器及存储在所述存储器上并可在所述处理器上运行的光通道性能劣化智能预警程序，所述光通道性能劣化智能预警程序配置为实现如权利要求上文所述的光通道性能劣化智能预警方法的步骤。
第四方面，本发明还提出一种存储介质，所述存储介质上存储有光通道性能劣化智能预警程序，所述光通道性能劣化智能预警程序被处理器执行时实现如上文所述的光通道性能劣化智能预警方法的步骤。
本发明提出的光通道性能劣化智能预警方法，通过获取电信传送网中待监测光通道的性能数据，根据所述性能数据构建所述待监测光通道的关键绩效指标KPI性能预测模型；根据KPI性能预测模型对所述待监测光通道的性能进行周期趋势预测，获得性能趋势预测结果；根据性能趋势预测结果判断所述待监测光通道是否出现性能劣化，在出现性能劣化时，进行预警处理， 能够对传送网光通道KPI性能指标进行针对性建模，保障了模型在工程长期应用中始终保持较好的预测精度；同时基于模型对光通道KPI性能指标进行长短期性能与趋势的预测和分析，实现对光通道性能劣化的提前感知和自动化预警，主动提醒运维人员提前对劣化光通道进行人工干预，防患于未然。
附图说明
图1为本发明实施例方案涉及的硬件运行环境的设备结构示意图；
图2为本发明光通道性能劣化智能预警方法第一实施例的流程示意图；
图3为本发明光通道性能劣化智能预警方法第二实施例的流程示意图；
图4为本发明光通道性能劣化智能预警方法第三实施例的流程示意图；
图5为本发明光通道性能劣化智能预警方法第四实施例的流程示意图；
图6为本发明光通道性能劣化智能预警方法第五实施例的流程示意图；
图7为本发明光通道性能劣化智能预警方法第六实施例的流程示意图；
图8为本发明光通道性能劣化智能预警方法第七实施例的流程示意图；
图9为本发明光通道性能劣化智能预警方法第八实施例的流程示意图；
图10为本发明光通道性能劣化智能预警方法第九实施例的流程示意图；
图11为本发明光通道性能劣化智能预警方法第十实施例的流程示意图；
图12为本发明光通道性能劣化智能预警装置第一实施例的功能模块图。
本发明目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。
具体实施方式
应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用于限定本发明。
本发明实施例的解决方案主要是：通过获取电信传送网中待监测光通道的性能数据，根据所述性能数据构建所述待监测光通道的关键绩效指标KPI性能预测模型；根据KPI性能预测模型对所述待监测光通道的性能进行周期趋势预测，获得性能趋势预测结果；根据性能趋势预测结果判断所述待监测光通道是否出现性能劣化，在出现性能劣化时，进行预警处理，能够对传送网光通道KPI性能指标进行针对性建模，保障了模型在工程长期应用中始终 保持较好的预测精度；同时基于模型对光通道KPI性能指标进行长短期性能与趋势的预测和分析，实现对光通道性能劣化的提前感知和自动化预警，主动提醒运维人员提前对劣化光通道进行人工干预，防患于未然，解决了现有技术中不能针对性的确定光通道性能劣化，且缺乏对光通道性能劣化的预测能力的技术问题。
参照图1，图1为本发明实施例方案涉及的硬件运行环境的设备结构示意图。
如图1所示，该设备可以包括：处理器1001，例如CPU，通信总线1002、用户端接口1003，网络接口1004，存储器1005。其中，通信总线1002用于实现这些组件之间的连接通信。用户端接口1003可以包括显示屏(Display)、输入单元比如键盘(Keyboard)，可选用户端接口1003还可以包括标准的有线接口、无线接口。网络接口1004可选的可以包括标准的有线接口、无线接口(如Wi-Fi接口)。存储器1005可以是高速RAM存储器，也可以是稳定的存储器(Non-Volatile Memory)，例如磁盘存储器。存储器1005可选的还可以是独立于前述处理器1001的存储装置。
本领域技术人员可以理解，图1中示出的设备结构并不构成对该设备的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。
如图1所示，作为一种存储介质的存储器1005中可以包括操作装置、网络通信模块、用户端接口模块以及光通道性能劣化智能预警程序。
本发明设备通过处理器1001调用存储器1005中存储的光通道性能劣化智能预警程序，并执行本发明实施例提供的光通道性能劣化智能预警方法。
基于上述硬件结构，提出本发明光通道性能劣化智能预警方法实施例。
参照图2，图2为本发明光通道性能劣化智能预警方法第一实施例的流程示意图。
在第一实施例中，所述光通道性能劣化智能预警方法包括以下步骤：
步骤S10、获取电信传送网中待监测光通道的性能数据，根据所述性能数据构建所述待监测光通道的关键绩效指标KPI性能预测模型。
需要说明的是，所述待监测光通道为电信传送网中需要进行监测的当前光通道，所述待监测光通道可以为一个，也可以为多个，本实施例对此不加以限制；所述性能数据为能够表明所述待监测光通道的当前光通道性能的数据，所述性能数据可以是获取一段时间内的性能参数作为性能数据，例如获取所述待监测光通道15分钟粒度的原始性能数据，当然也可以通过其他方式采集所述待监测光通道的性能数据，本实施例对此不加以限制；所述性能参数可以为光通道的误码数据，也可以是光功率数据，还可以是光信噪比(Optical Signal Noise Ratio，OSNR)数据，当然还可以是其他类型的数据，或者多个数据的组合，本实施例对此不加以限制；在获取了所述性能数据后，可以根据所述性能数据构建所述待监测光通道的关键绩效指标(Key Performance Indicator，KPI)性能预测模型。
步骤S20、根据KPI性能预测模型对所述待监测光通道的性能进行周期趋势预测，获得性能趋势预测结果。
应当理解的是，在获得了KPI性能预测模型后，可以通过所述KPI性能预测模型获得所述待监测光通道的在未来的某段时间的性能趋势，周期性趋势预测可以是长周期性能预测，也可以是短周期性能预测，具体根据实际使用需求确定，在进行了性能的周期趋势预测后，能够获得对应的性能趋势预测结果，即在未来某段时间的性能变化趋势。
步骤S30、根据性能趋势预测结果判断所述待监测光通道是否出现性能劣化，在出现性能劣化时，进行预警处理。
可以理解的是，在获得了性能趋势预测结果后，可以对所述性能趋势预测结果进行分析，判断其性能趋势是否出现性能趋势劣化，如果出现劣化则进行相应的预警处理，所述预警处理为自动化预警，即根据性能趋势劣化的程度采取不同的预警通知策略。
本实施例通过上述方案，通过获取电信传送网中待监测光通道的性能数据，根据所述性能数据构建所述待监测光通道的关键绩效指标KPI性能预测模型；根据KPI性能预测模型对所述待监测光通道的性能进行周期趋势预测，获得性能趋势预测结果；根据性能趋势预测结果判断所述待监测光通道是否出现性能劣化，在出现性能劣化时，进行预警处理，能够对传送网光通道KPI性能指标进行针对性建模，保障了模型在工程长期应用中始终保持较好的预 测精度；同时基于模型对光通道KPI性能指标进行长短期性能与趋势的预测和分析，实现对光通道性能劣化的提前感知和自动化预警，主动提醒运维人员提前对劣化光通道进行人工干预，防患于未然。
进一步地，图3为本发明光通道性能劣化智能预警方法第二实施例的流程示意图，如图3所示，基于第一实施例提出本发明光通道性能劣化智能预警方法第二实施例，在本实施例中，所述步骤S10具体包括以下步骤：
步骤S11、获取电信传送网中若干待监测光通道的性能数据。
需要说明的是，不同的待监测光通道有不同类型的性能数据，通过获取电信传送网中若干待监测光通道的性能数据，能够作为后续确定对应的模型算法的依据；一般可以通过网管性能上报通道周期性的获取待监测光通道的性能数据，例如，通过网管性能上报通道每15分钟获取所要监测光通道的当前15分钟性能数据包，从中解析出待监测光通道的性能数据，所述性能数据包括但不限于：光通道误码率、源/宿节点收发光功率、光通道信噪比等。
步骤S12、根据各性能数据确定各待监测光通道的各性能预测目标，根据各性能预测目标从预设数据库中选取对应的若干目标模型算法。
可以理解的是，根据各性能数据确定各待监测光通道的各性能预测目标，例如误码性能数据对应的性能预测目标为信道数据流，收发光功率数据对应的性能预测目标为光信号输出和接收频率，OSNR数据对应的性能预测目标为光有效带宽为0.1nm内光信号功率和噪声功率等，本实施例对此不加以限制；不同的性能预测目标对应有不同的模型算法，模型算法可以预先存储在预设数据库中，所述模型算法可以选择基于自回归预测的机器学习模型，例如随机森林模型、岭回归模型、梯度增强回归树(Gradient Boosting Regression Trees，GBRT)模型以及支持向量回归模型(Support Vector Regression，SVR)等；也可以是基于时序预测的深度学习模型，例如长短期记忆网络(Long Short-Term Memory，LSTM)以及门控循环单元(Gated Recurrent Unit，GRU)等，本实施例对此不加以限制。
步骤S13、根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测模型。
应当理解的是，在确定了各性能预测目标和各目标模型算法后，可以根 据各性能预测目标和各目标模型算法构件各自对应的KPI性能预测模型。
本实施例通过上述方案，通过获取电信传送网中若干待监测光通道的性能数据；根据各性能数据确定各待监测光通道的各性能预测目标，根据各性能预测目标从预设数据库中选取对应的若干目标模型算法；根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测模型，能够对待监测光通道进行针对性的KPI性能预测模型的建模，保障了模型的预测精度的长期有效。
进一步地，图4为本发明光通道性能劣化智能预警方法第三实施例的流程示意图，如图4所示，基于第二实施例提出本发明光通道性能劣化智能预警方法第三实施例，在本实施例中，所述步骤S13具体包括以下步骤：
步骤S131、根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测候选基线模型。
需要说明的是，所述预测候选基线模型为针对不同的预测目标和不同的模型算法组合构建的不同的对待监测光通道的关键绩效指标KPI性能记性预测的模型，不同的模型会有不同的技术结果，而计算同一预测目标，不同的算法会对应不同的预测精度。
步骤S132、获取各待监测光通道的历史性能数据，根据所述历史性能数据从各KPI性能预测候选基线模型中选取最优基线模型。
可以理解的是，所述历史性能数据为记录了各待监测光通道在当前时刻之前的性能数据，可以是之前所有的性能数据，也可以是之前某一段时间的历史性能数据，例如过去30天的光通道历史数据，本实施例对此不加以限制；通过所述历史性能数据对各KPI性能预测候选基线模型进行训练，可以获得最优的基线模型，一般是选取预测精度最高的模型作为最优基线模型，当然也可以是根据其他方式确定最优基线模型，例如将预测速度最快的基线模型作为最优基线模型，本实施例对此不加以限制。
步骤S133、获取若干待监测光通道的通道参数。
需要说明的是，所述待监测光通道的通道参数为所述待监测光通道的基础数据，一般可以通过传送网管理系统所提供的北向接口获取，也可以通过私有协议获取，当然也可以通过其他方式获取所述通道数据，本实施例对此 不加以限制；所述通道数据包括但不限于：光通道名称、速率、波长、源宿网元以及中间路由等；也可以包括网管为光通道各性能指标所预配置的静态阈值；当然还可以包括其他数据，本实施例对此不加以限制。
在具体实现中，获取若干待监测光通道的通道参数之后，一般将所述通道数据保存到光通道基础数据库中，以便后续应用时方便查找；可以保持光通道基础数据与网管系统同步；同步策略可以采用固定周期同步或实时增量同步方式；前者采用每隔一段时间从网管系统获取最新的光通道数据并进行同步，后者采用监听网管系统配置变更事件，每当网管系统发生光通道基础数据变化时才进行增量同步。
步骤S134、根据各通道参数将各待监测光通道划分为相同路由不同波长的光通道组。
可以理解的是，通过各通道参数可以将各待监测光通道进行分组，即比较各待监测光通道的路由，将相同路由不同波长的光通道划分为一个光通道组，分组有助于后续针对性的进行分组训练，提高光通道的KPI性能预测的精度和预测速度。
步骤S135、基于所述最优基线模型对各光通道组进行分组训练，获得各待监测光通道的KPI性能预测模型。
应当理解的是，基于所述最优基线模型对各光通道组进行分组训练，即针对每一组光通道分别选取对应的最优基线模型进行分组预测模型训练，在实际操作中，一般从光通道历史性能库中获取本组光通道相关KPI性能的最新数据作为本组训练数据，由于相同路由的光通道一般会具有相似的性能劣化规律，因此，可以通过对光通道进行模型分组泛化来一方面减少光通道性能预测所需的模型总量和模型训练计算量，另一方面可同时继续保持较高的模型预测精度。
需要说明的是，基于所述最优基线模型对各光通道组进行分组训练之后，可以将分组训练训练好的分组泛化模型作为各待监测光通道的KPI性能预测模型，即训练好的光通道KPI性能预测的分组泛化模型作为该组光通道组近段时间将上线使用的正式模型，并将该模型及其模型相关信息保存于本装置的光通道模型库中；所述KPI性能预测模型的模型基本信息包括但不限于：模型ID、模型分类、模型用途、模型算法、最后训练日期、训练精度、相关 基线模型版本以及模型适用的光通道ID信息等，当然也可以包含更多或更少的信息，本实施例对此不加以限制。
在具体实现中，可以对训练好的光通道KPI性能预测的分组泛化模型的后续的实际预测精度进行监测；若后续监测中发现：模型的预测精度明显下降时，例如，准确率低于80％；或相关光通道发生路由或配置变更时；此时需要重新启动对相关光通道KPI性能分组泛化模型的再次训练，以再提升模型预测精度。
本实施例通过上述方案，通过根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测候选基线模型；获取各待监测光通道的历史性能数据，根据所述历史性能数据从各KPI性能预测候选基线模型中选取最优基线模型；获取若干待监测光通道的通道参数；根据各通道参数将各待监测光通道划分为相同路由不同波长的光通道组；基于所述最优基线模型对各光通道组进行分组训练，获得各待监测光通道的KPI性能预测模型，能够减少光通道性能预测所需的模型总量和模型训练计算量，进一步提高了模型的预测精度。
进一步地，图5为本发明光通道性能劣化智能预警方法第四实施例的流程示意图，如图5所示，基于第三实施例提出本发明光通道性能劣化智能预警方法第四实施例，在本实施例中，所述步骤S132具体包括以下步骤：
步骤S1321、获取各待监测光通道的历史性能数据。
需要说明的是，所述历史性能数据为在过去一段时间内的待监测光通道产生的性能数据记录，也可以为待监测光通道在当前时刻之前的所有性能数据记录，本实施例对此不加以限制。
步骤S1322、根据所述历史性能数据对各KPI性能预测候选基线模型进行训练，获得各KPI性能预测候选基线模型的训练精度。
应当理解的是，根据所述历史性能数据对各KPI性能预测候选基线模型进行训练，及将历史性能数据代入至各KPI性能预测候选基线模型中，将获得的预测值与真实值比较计算后获得各KPI性能预测候选基线模型的训练精度。
在具体实现中，KPI性能预测候选基线模型可以是短周期预测模型和长 周期预测模型；短周期预测模型可以对光通道KPI性能的15分钟粒度性能进行日趋势预测，模型目标是从过去N天(N一般取≥10天)的光通道历史性能数据(15分钟粒度)预测第N+1天所有15分钟粒度性能值；而长周期预测模型可以对光通道KPI性能的M日平均日趋势进行预测，M一般取3日或5日(同上)，以便既能较好的反映光通道的趋势性，又对趋势的变化具有较好的灵敏性，长周期预测模型目标是从过去N天(N一般取≥30天)的光通道历史数据预测N+M天的M日均线的性能趋势值，即M日内的平均日均值、平均日最大值以及平均日最小值。
步骤S1323、从各KPI性能预测候选基线模型中选取训练精度最高的KPI性能预测候选基线模型作为最优基线模型。
可以理解的是，根据模型算法的不同和KPI指标的不同，每一大类模型可细分为多个子类模型；对所有候选模型的训练可以选取不同工程3个月以上的光通道历史性能数据进行训练。
在具体实现中，模型的训练精度(即准确率)统一按如下进行计算：
模型偏差率＝|预测值-真实值|/真实值
模型准确率＝偏差率小于20％的光通道数/所有预测光通道数
选取训练精度最好的候选模型作为光通道某类KPI性能预测的基线模型，并且可以将该基线模型以及模型基本信息保存于的光通道模型库中，所述模型基本信息包括但不限于基线模型ID、模型分类、模型算法、模型版本、更新时间及摘要描述等，当然也可以包括更多或更少的信息，本实施例对此不加以限制。
本实施例通过上述方案，通过获取各待监测光通道的历史性能数据；根据所述历史性能数据对各KPI性能预测候选基线模型进行训练，获得各KPI性能预测候选基线模型的训练精度；从各KPI性能预测候选基线模型中选取训练精度最高的KPI性能预测候选基线模型作为最优基线模型，能够保障性能预测的精确度，并且保证模型精度的长期有效，为后续光通道性能异常检测提供检测基础。
进一步地，图6为本发明光通道性能劣化智能预警方法第五实施例的流程示意图，如图6所示，基于第一实施例提出本发明光通道性能劣化智能预 警方法第五实施例，在本实施例中，所述步骤S20具体包括以下步骤：
步骤S21、获取KPI性能预测模型的模型类别和模型用途。
需要说明的是，所述模型类别为根据所述KPI性能预测模型的类型进行划分的类别，根据所述模型类别可以确定所述KPI性能预测模型是短周期预测模型或是长周期预测模型；所述模型用途为根据所述KPI性能预测模型的预测对象进行划分的用途，根据所述模型用途可以确定所述KPI性能预测模型是误码预测模型，还是收发光功率预测模型，还是信噪比预测模型，或者其他类型的性能参数预测模型，本实施例对此不加以限制。
步骤S22、根据所述模型类别和所述模型用途从各待监测光通道的历史性能库中读取所需周期的目标性能数据。
可以理解的是，通过所述模型类别和所述模型用途能够从所述历史性能库中获取匹配的性能数据，即与所述模型类别和所述模型用途对应的所需周期内的目标性能数据。
在具体实现中，对于每条光通道KPI性能指标的短周期性能预测来说，一般通过查询光通道模型库，所述光通道模型库为用于存储各类光通道模型的数据库，所述光通道模型库可以在本地存储，也可以是在云端存储，本实施例对此不加以限制；获取相关光通道本类KPI性能预测所对应的光通道分组泛化模型，具体是通过模型分类、模型用途以及模型所适用的光通道ID与本光通道ID是否匹配进行模型选择；根据所获取的光通道分组预测模型的模型类别和用途的不同，从本装置的光通道历史性能库中读取该模型所需的过去N天(N一般取10天)预设时间的原始性能数据，一般可以为15分钟的原始性能数据，即作为目标性能数据；当然也可以为其他预设时间，本实施例对此不加以限制；对于长周期性能预测来说，与短周期性能预测类似，也是通过光通道模型库确定匹配的模型，根据所获取的光通道分组泛化模型的模型类别和用途不同，从光通道历史性能库中读取该模型所需的过去N天(N一般取30天)M日聚合数据，以此作为目标性能数据。
步骤S23、将所述目标性能数据输入至所述KPI性能预测模型中，获得所述所需周期的性能趋势预测结果。
应当理解的是，将所述目标性能数据输入至所述KPI性能预测模型中，获得的输出结果即为所需周期的性能趋势预测结果；对于短周期性能预测来 说，通过动态加载并允许所述KPI性能预测模型，模型运行结束后获取模型的输出结果，该结果即为相关光通道第N+1天预设时间的性能预测值；对于长周期性能预测来说，模型运行结束后获取模型的输出结果，该结果即为相关光通道未来M日KPI性能的日均趋势预测值；在获得了性能趋势预测结果后，一般会将相应的预测日期和结果一起保存在光通道预测数据库中，以便于后续进一步的分析计算。
本实施例通过上述方案，通过获取KPI性能预测模型的模型类别和模型用途；根据所述模型类别和所述模型用途从各待监测光通道的历史性能库中读取所需周期的目标性能数据；将所述目标性能数据输入至所述KPI性能预测模型中，获得所述所需周期的性能趋势预测结果，能够根据实际需求针对性的进行性能预测，提高了性能预测的灵活性和精确度。
进一步地，图7为本发明光通道性能劣化智能预警方法第六实施例的流程示意图，如图7所示，基于第一实施例提出本发明光通道性能劣化智能预警方法第六实施例，在本实施例中，所述步骤S30具体包括以下步骤：
步骤S31、从所述性能趋势预测结果中获得各个时段的分时动态阈值。
需要说明的是，所述分时动态阈值为各个时段的性能趋势预测值对应的预设阈值，各个时段的分时动态阈值一般是不同的，动态变化的。
步骤S32、将所述性能趋势预测结果与所述分时动态阈值比较，判断所述待监测光通道是否出现当前性能异常。
可以理解的是，通过将同一时刻的性能趋势预测结果和对应的分时动态阈值比较，能够根据比较结果确定是否出现当前性能异常；在实际操作中一般会对所述待监测光通道进行实时异常检测，从而提升数据处理效率以及对光通道性能预警的时效性。
步骤S33、从所述性能趋势预测结果中获得趋势预测值，根据所述趋势预测值是否进入了预设性能劣化风险区判断所述待监测光通道是否出现趋势异常。
应当理解的是，所述性能趋势预测结果中有未来某个时间段的趋势性能预测，所述预设性能劣化风险区为预先设置的辅助性能趋势劣化判断的动态风险区，通过所述趋势预测值是否进入了预设性能劣化风险区能够判断所述 待监测光通道是否出现趋势异常。
在具体实现中，从所述性能趋势预测结果中获得趋势预测值，一般是通过从历史性能数据库中获取相关光通道KPI性能指标过去N天的M日预聚合数据作为趋势预测值，所述预设性能劣化风险区可以分为上线中风险区、上线高风险区、下线中风险区和下线高风险区；当然也可以设置为其他类型的性能劣化风险区，所述预设性能劣化风险区一般通过趋势预测值计算获得以上线中风险区、上线高风险区、下线中风险区和下线高风险区为例：
R1＝过去N天的实际KPI性能平均日均值
R2＝过去N天的实际KPI性能平均日最大值
R3＝过去N天的实际KPI性能平均日最小值
R4＝Min(R1+(R2-R1)×2，上线静态阈值)
R5＝Max(R1-(R1-R3)×2，下线静态阈值)
上线中风险区U_Risk_0＝R1+(R4-R1)×70％
上线高风险区U_Risk_1＝R1+(R4-R1)×90％
下线中风险区D_Risk_0＝R1-(R1-R5)×70％
下线高风险区D_Risk_1＝R1-(R1-R5)×90％
其中，对于实际KPI性能平均日最大值、平均日最小值和平均日均值的计算需剔除掉异常点，对所监控光通道各KPI性能进行M日趋势预测值检测，判定其是否存在趋势异常；趋势异常的判定分为如下四类：
A类异常：判断未来M天光通道KPI性能的平均日最大值P max_i是否进入上限中风险区、且未来M天光通道KPI性能的平均日均线P avg_i是否持续上升。
判定公式：P max_i≥U_Risk_0&&P avg_i＜P avg_i+1＜P avg_i+m
B类异常：判断未来M天光通道KPI性能的平均日最小值P min_i是否进入下限中风险区、且未来M天光通道KPI性能的平均日均线P avg_i是否持续下降。
判定公式：P min_i≤D_Risk_0&&P avg_i＞P avg_i+1＞P avg_i+m
C类异常：判断未来M天光通道KPI性能的平均日最大值P max_i是否进入上限高风险区。
判定公式：P max_i≥U_Risk_1
D类异常：判断未来M天光通道KPI性能的平均日最大值P min_i是否进入下限高风险区。
判定公式：P min_i≤D_Risk_1
一般的，在对光通道KPI性能趋势检测中，对误码率趋势进行A类和C类异常的判定检测，对收发光功率趋势进行A类、B类、C类、D类异常的判定检测，对OSNR趋势进行B类和D类异常的判定检测；
并且，对于出现的趋势异常，通常会记录光通道性能趋势异常事件，并保存于光通道性能异常记录中；光通道性能趋势异常记录以下信息：相关KPI指标、异常分类(A/B/C/D类异常)、检测日期、相关光通道ID与名称、以及异常判定所使用的M日趋势预测值和风险判定值等信息；当然还可以包括更多或更少的信息，本实施例对此不加以限制。
步骤S34、在出现当前性能异常和/或趋势异常时，判定所述待监测光通道出现性能劣化。
可以理解的是，当出现了当前性能异常或趋势异常中的任一一种时，即可判定待监测光通道出现性能趋势劣化。
步骤S35、在所述待监测光通道出现性能劣化时，进行预警处理。
应当理解的是，在所述待监测光通道出现性能劣化时，会根据性能劣化的不同程度制定不同的预警处理方案进行预警。
本实施例通过上述方案，通过从所述性能趋势预测结果中获得各个时段的分时动态阈值；将所述性能趋势预测结果与所述分时动态阈值比较，判断所述待监测光通道是否出现当前性能异常；从所述性能趋势预测结果中获得趋势预测值，根据所述趋势预测值是否进入了预设性能劣化风险区判断所述待监测光通道是否出现趋势异常；在出现当前性能异常和/或趋势异常时，判定所述待监测光通道出现性能劣化；在所述待监测光通道出现性能劣化时，进行预警处理，能够对待监测光通道中的异常进行全面性的检测，从而提高了待监测光通道的性能检测精确度，实现对光通道性能劣化的提前感知和自动化预警。
进一步地，图8为本发明光通道性能劣化智能预警方法第七实施例的流程示意图，如图8所示，基于第六实施例提出本发明光通道性能劣化智能预 警方法第七实施例，在本实施例中，所述步骤S31具体包括以下步骤：
步骤S311、从所述性能趋势预测结果中获得各个时段的所有本时段预测值。
需要说明的是，所述性能趋势预测结果中对应有各个时间段的性能预测值，所述本时段预测值，即为预设时段的各时段的本时段内的预测值。
步骤S312、从所述所有本时段预测值中确定最大预测值和最小预测值。
可以理解的是，通过分析和筛选，能够从所有的本时段预测值张工确定最大的预测值和最小的预测值。
步骤S313、根据预设调整系数、所述最大预测值和所述最小预测值获得各个时段的分时动态阈值。
应当理解的是，所述预设调整系数为预先设置的计算分时动态阈值的调整系数，通过所述预设调整系数、所述最大预测值和所述最小预测值能够计算获得各个时段的分时动态阈值。
进一步地，所述步骤S313具体包括以下步骤：
根据预设调整系数、所述最大预测值和所述最小预测值利用下述公式获得各个时段的分时动态阈值：
G t＝G max+|G max-G min|×K
其中，G t为各个时段的分时动态阈值，G max为所述所有本时段预测值中的最大预测值，G min为所述所有本时段预测值中的最小预测值，K为预设调整系数。
在具体实现中，在完成光通道KPI性能预测后，基于预测值可以计算出该光通道的第N+1天的分时动态阈值，考虑到光通道性能数据变化特征和实际监控价值，分时时段定义为每H小时为一个时段，H一般取值为1/2/3/4小时即可，默认取2小时，调整系数K可以为1.0～2.0之间固定取值，初始值可取1.0，后续可根据光通道实际监控效果进行调整；当然也可以是其他数值，本实施例对此不加以限制；在获得了各个时段的分时动态阈值后，可以将其进行存储，可以生成光通道性能分时动态阈值表，该表可以在后续的光通道当前性能异常分析中使用。
本实施例通过上述方案，通过从所述性能趋势预测结果中获得各个时段的所有本时段预测值；从所述所有本时段预测值中确定最大预测值和最小预 测值；根据预设调整系数、所述最大预测值和所述最小预测值获得各个时段的分时动态阈值，能够分时段的设置动态阈值，进一步提升了性能预测的精确度，作为光通道性能劣化判断依据。
进一步地，图9为本发明光通道性能劣化智能预警方法第八实施例的流程示意图，如图9所示，基于第六实施例提出本发明光通道性能劣化智能预警方法第八实施例，在本实施例中，所述步骤S32具体包括以下步骤：
步骤S321、从所述性能趋势预测结果中获取当前时刻对应的当前性能数据，并从所述分时动态阈值中获取所述当前时刻的当前动态阈值。
需要说明的是，所述性能趋势预测结果中存储有各个时段的性能趋势预测数据，可以从中获取与所述当前时刻匹配的性能数据，并且可以从所述分时动态阈值中获得当前时刻对应的当前动态阈值。
步骤S322、判断所述当前性能数据是否为无效数据。
可以理解的是，所述无效数据一般包括性能值缺失、过大或过小值、非法零值或无效Null值等，当然也可以包括其他类型的无效数据，本实施例对此不加以限制；在实际操作中，一般通过检测所监控光通道当前15分钟粒度的KPI性能数据作为当前性能数据，判断所述当前性能数据是否为无效数据，如是无效数据则丢弃掉，不做任何处理。
步骤S323、在所述当前数据不为无效数据时，获取所述待监测光通道的光通道性能静态阈值。
需要说明的是，所述光通道性能静态阈值为预先设置光通道性能的固定的阈值；一般的，从光通道KPI性能分时动态阈值缓存表中，读取该条光通道当前时刻所对应的当前动态阈值，并从光性能基础数据中读取从网管中所获取的光通道性能静态阈值。
步骤S324、将所述当前性能数据分别与所述光通道性能静态阈值及所述当前动态阈值的比较。
可以理解的是，通过将所述当前性能数据和所述光通道性能静态阈值及所述当前动态阈值的比较，能够更加全面的判断所述待监测光通道未出现当前性能异常。
步骤S325、在所述当前性能数据不大于所述光通道性能静态阈值及所述 当前动态阈值时，判定所述待监测光通道未出现当前性能异常。
应当理解的是，在所述当前性能数据都不大于光通道性能静态阈值和当前动态阈值时，即可判定所述待监测光通道未出现当前性能异常。
步骤S326、在所述当前性能数据大于所述光通道性能静态阈值和/或所述当前动态阈值时，判定所述待监测光通道出现当前性能异常。
应当理解的是，在所述当前性能数据大于光通道性能静态阈值和当前动态阈值中的一个或两个时，即可判定所述待监测光通道出现当前性能异常。
在具体实现中，在确定了当前性能异常时，可以记录一条光通道当前性能异常事件，并保存于光通道性能库异常记录中；光通道当前性能异常事件记录以下信息：相关KPI性能指标、异常分类(静态越限、动态越限)、异常发生时间、相关光通道ID与名称、异常性能值、异常判定所使用的分时性能动态阈值和静态阈值等信息。
本实施例通过上述方案，通过从所述性能趋势预测结果中获取当前时刻对应的当前性能数据，并从所述分时动态阈值中获取所述当前时刻的当前动态阈值；判断所述当前性能数据是否为无效数据；在所述当前数据不为无效数据时，获取所述待监测光通道的光通道性能静态阈值；将所述当前性能数据分别与所述光通道性能静态阈值及所述当前动态阈值的比较；在所述当前性能数据不大于所述光通道性能静态阈值及所述当前动态阈值时，判定所述待监测光通道未出现当前性能异常；在所述当前性能数据大于所述光通道性能静态阈值和/或所述当前动态阈值时，判定所述待监测光通道出现当前性能异常，能够更加全面的判断当前性能异常，进一步提高了光通道性能检测的精确度。
进一步地，图10为本发明光通道性能劣化智能预警方法第九实施例的流程示意图，如图10所示，基于第六实施例提出本发明光通道性能劣化智能预警方法第九实施例，在本实施例中，所述步骤S35具体包括以下步骤：
步骤S351、在所述待监测光通道出现性能趋势劣化时，获取当前性能异常和/或趋势异常的异常次数。
需要说明的是，所述异常次数为所述待监测光通道出现性能趋势劣化时，当前性能异常和趋势异常的总异常次数，当然有可能出现只有一种异常的情 况，此时所述异常次数为该异常的总异常次数。
步骤S352、根据所述异常次数从预设预警策略中确定对应的目标预警处理策略，根据所述目标预警处理策略进行预警通知。
可以理解的是，不同的异常次数对应不同的预警处理策略，所述预设预警策略中存储有不同异常次数对应不同的预警处理策略，根据目标预警处理策略可以针对性的进行预警处理操作，实现对光通道性能劣化的提前感知和自动化预警，提醒运维人员提前对劣化光通道进行人工干预，防患于未然。
在具体实现中，所述预设预警策略可以包括两个部分：一是对光通道KPI性能异常事件何时触发、以及触发何种级别预警的触发策略，另一是对光通道性能异常触发预警后的通知策略；基本的预警触发策略定义如下：
1)对首次发生某KPI性能当前异常的光通道进行一级预警；
2)对同时发生多个KPI性能当前异常的光通道进行二级预警；
3)对持续多次发生(如：3次)某KPI性能当前异常的光通道进行二级预警；
4)对首次发生某KPI性能趋势异常的光通道进行二级预警；
5)对同时发生多类KPI性能趋势异常的光通道进行三级预警；
6)对持续多次发生(如：3次)某KPI性能趋势异常的光通道进行三级预警；
基本的预警通知策略包括：邮件通知、短信通知、系统通知以及组合通知，当然还可以通过其他方式，例如电话通知，声音光线告警等，本实施例对此不加以限制；一般的，在收到光通道KPI性能当前异常或趋势异常的通知后，自动启动预警监测工作；从异常事件库中分析统计与当前所上报异常的光通道相关的近期异常记录，并根据上一步所预定义的预警触发策略判断当前光通道近期累计异常是否符合策略所定义的触发条件，对于符合预警触发条件的光通道性能与趋势劣化事件，根据预警通知策略进行自动预警通知。
本实施例通过上述方案，通过在所述待监测光通道出现性能趋势劣化时，获取当前性能异常和/或趋势异常的异常次数；根据所述异常次数从预设预警策略中确定对应的目标预警处理策略，根据所述目标预警处理策略进行预警通知，能够针对性的进行预警处理操作，实现对光通道性能劣化的提前感知和自动化预警，主动提醒运维人员提前对劣化光通道进行人工干预，防患于 未然。
进一步地，图11为本发明光通道性能劣化智能预警方法第十实施例的流程示意图，如图11所示，基于第二实施例提出本发明光通道性能劣化智能预警方法第十实施例，在本实施例中，所述步骤S11具体包括以下步骤：
步骤S111、对所述性能数据进行无效数据检查。
需要说明的是，通过对所述性能数据进行无效数据检查，能够排查出无效数据，进一步提升性能检测的精确度；所述无效数据一般包括性能值缺失、过大或过小值、非法零值或无效Null值等，当然也可以包括其他类型的无效数据，本实施例对此不加以限制。
步骤S112、在所述性能数据中出现目标无效数据时，获取所述目标无效数据的上一预设时刻或下一预设时刻的有效性能值。
应当理解的是，在所述性能数据中出现目标无效数据时，通过获取所述目标无效数据的上一预设时刻或下一预设时刻的有效性能值，能够对无效数据进行代替，从而保证预测的准确性。
在具体实现中，还可以对对原始性能相关的日期时间字段进行规范化处理，统一数据精度，并转换成开发语言易于处理的时间数据类型，并将处理后的光通道15分钟KPI性能数据保存到光通道历史性能数据库中，从而进一步提高性能检测的精确度，并提升检测速度和效率。
步骤S113、将所述目标无效数据替换为所述有效性能值，将替换后的性能数据作为新的性能数据。
可以理解的是，通过将有效性能值替换掉所述目标无效数据，能够进一步保证性能预测的准确性，保障了模型的预测精度的长期有效。
在具体实现中，可以每日定时对光通道15分钟性能数据进行预聚合计算，预聚合采用滑动时间窗口的方式，基于原始15分钟性能数据，计算出每条光通道KPI性能在过去M天的平均数据；为既能体现光通道性能的趋势性、同时又对趋势变化保持较好的敏感性，M一般取3天或5天，默认取5天；需预聚合计算的数据包括：过去M天内的平均日均值、平均日最大值、平均日最小值；需要注意的是，预聚合计算时需剔除掉一些偶发的异常点，通常这类偶发异常点是一个个孤点，该点的性能值明显超出或明显低出其它点几 个数量级；由于这种偶发异常往往是由于突发的网络故障或突发的人为操作所致，是难以预测的，同时也不能反映光通道正常的性能劣化趋势，因此需要剔除掉，避免带入到后续对趋势的分析预测中，影响预测的准确性；对这些偶发异常点的预聚合处理可统一按上一个或下一个有效点的性能值代替即可；性能预聚合的计算结果同步保存到光通道历史性能数据库中，以便于后续的分析计算。
本实施例通过上述方案，通过对所述性能数据进行无效数据检查；
在所述性能数据中出现目标无效数据时，获取所述目标无效数据的上一预设时刻或下一预设时刻的有效性能值；
将所述目标无效数据替换为所述有效性能值，将替换后的性能数据作为新的性能数据，能够进一步保证性能预测的准确性，保障了模型的预测精度的长期有效，并提升检测速度和效率。
相应地，本发明进一步提供一种光通道性能劣化智能预警装置。
参照图12，图12为本发明光通道性能劣化智能预警装置第一实施例的功能模块图。
本发明光通道性能劣化智能预警装置第一实施例中，该光通道性能劣化智能预警装置包括：
模型构建模块10，用于获取电信传送网中待监测光通道的性能数据，根据所述性能数据构建所述待监测光通道的关键绩效指标KPI性能预测模型。
预测模块20，用于根据KPI性能预测模型对所述待监测光通道的性能进行周期趋势预测，获得性能趋势预测结果。
预警模块30，用于根据性能趋势预测结果判断所述待监测光通道是否出现性能劣化，在出现性能劣化时，进行预警处理。
进一步地，所述模型构建模块10包括：
性能获取模块，用于获取电信传送网中若干待监测光通道的性能数据；
算法选取模块，用于根据各性能数据确定各待监测光通道的各性能预测目标，根据各性能预测目标从预设数据库中选取对应的若干目标模型算法；
模型生成模块，用于根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测模型。
进一步地，所述模型生成模块包括：
基线模型构建模块，用于根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测候选基线模型；
选取模块，用于获取各待监测光通道的历史性能数据，根据所述历史性能数据从各KPI性能预测候选基线模型中选取最优基线模型；
参数获取模块，用于获取若干待监测光通道的通道参数；
划分模块，用于根据各通道参数将各待监测光通道划分为相同路由不同波长的光通道组；
分组训练模块，用于基于所述最优基线模型对各光通道组进行分组训练，获得各待监测光通道的KPI性能预测模型。
其中，光通道性能劣化智能预警装置的各个功能模块实现的步骤可参照本发明光通道性能劣化智能预警方法的各个实施例，此处不再赘述。
此外，本发明实施例还提出一种存储介质，所述存储介质可以是计算机可读非易失性存储介质，当然也可以是其他类型的存储介质，本实施例对此不加以限制；所述存储介质上存储有光通道性能劣化智能预警程序，所述光通道性能劣化智能预警程序被处理器执行时实现如上本发明实施例提供的光通道性能劣化智能预警的实施方法。
上述本发明实施例序号仅仅为了描述，不代表实施例的优劣。
以上仅为本发明的优选实施例，并非因此限制本发明的专利范围，凡是利用本发明说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本发明的专利保护范围内。

Claims (16)

  1. 一种光通道性能劣化智能预警方法，其特征在于，所述光通道性能劣化智能预警方法，包括：
    获取电信传送网中待监测光通道的性能数据，根据所述性能数据构建所述待监测光通道的关键绩效指标KPI性能预测模型；
    根据KPI性能预测模型对所述待监测光通道的性能进行周期趋势预测，获得性能趋势预测结果；
    根据性能趋势预测结果判断所述待监测光通道是否出现性能劣化，在出现性能劣化时，进行预警处理。
  2. 如权利要求1所述的光通道性能劣化智能预警方法，其特征在于，所述获取电信传送网中待监测光通道的性能数据，根据所述性能数据构建所述待监测光通道的关键绩效指标KPI性能预测模型，包括：
    获取电信传送网中若干待监测光通道的性能数据；
    根据各性能数据确定各待监测光通道的各性能预测目标，根据各性能预测目标从预设数据库中选取对应的若干目标模型算法；
    根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测模型。
  3. 如权利要求2所述的光通道性能劣化智能预警方法，其特征在于，所述根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测模型，包括：
    根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测候选基线模型；
    获取各待监测光通道的历史性能数据，根据所述历史性能数据从各KPI性能预测候选基线模型中选取最优基线模型；
    获取若干待监测光通道的通道参数；
    根据各通道参数将各待监测光通道划分为相同路由不同波长的光通道组；
    基于所述最优基线模型对各光通道组进行分组训练，获得各待监测光通道的KPI性能预测模型。
  4. 如权利要求3所述的光通道性能劣化智能预警方法，其特征在于，所述获取各待监测光通道的历史性能数据，根据所述历史性能数据从各KPI性能预测候选基线模型中选取最优基线模型，包括：
    获取各待监测光通道的历史性能数据；
    根据所述历史性能数据对各KPI性能预测候选基线模型进行训练，获得各KPI性能预测候选基线模型的训练精度；
    从各KPI性能预测候选基线模型中选取训练精度最高的KPI性能预测候选基线模型作为最优基线模型。
  5. 如权利要求1所述的光通道性能劣化智能预警方法，其特征在于，所述根据KPI性能预测模型对所述待监测光通道的性能进行周期趋势预测，获得性能趋势预测结果，包括：
    获取KPI性能预测模型的模型类别和模型用途；
    根据所述模型类别和所述模型用途从各待监测光通道的历史性能库中读取所需周期的目标性能数据；
    将所述目标性能数据输入至所述KPI性能预测模型中，获得所述所需周期的性能趋势预测结果。
  6. 如权利要求1所述的光通道性能劣化智能预警方法，其特征在于，所述根据性能趋势预测结果判断所述待监测光通道是否出现性能劣化，在出现性能劣化时，进行预警处理，包括：
    从所述性能趋势预测结果中获得各个时段的分时动态阈值；
    将所述性能趋势预测结果与所述分时动态阈值比较，判断所述待监测光通道是否出现当前性能异常；
    从所述性能趋势预测结果中获得趋势预测值，根据所述趋势预测值是否进入了预设性能劣化风险区判断所述待监测光通道是否出现趋势异常；
    在出现当前性能异常和/或趋势异常时，判定所述待监测光通道出现性能 劣化；
    在所述待监测光通道出现性能劣化时，进行预警处理。
  7. 如权利要求6所述的光通道性能劣化智能预警方法，其特征在于，所述从所述性能趋势预测结果中获得各个时段的分时动态阈值，包括：
    从所述性能趋势预测结果中获得各个时段的所有本时段预测值；
    从所述所有本时段预测值中确定最大预测值和最小预测值；
    根据预设调整系数、所述最大预测值和所述最小预测值获得各个时段的分时动态阈值。
  8. 如权利要求7所述的光通道性能劣化智能预警方法，其特征在于，根据预设调整系数、所述最大预测值和所述最小预测值利用下述公式获得各个时段的分时动态阈值：
    G t＝G max+|G max-G min|×K
    其中，G t为各个时段的分时动态阈值，G max为所述所有本时段预测值中的最大预测值，G min为所述所有本时段预测值中的最小预测值，K为预设调整系数。
  9. 如权利要求6所述的光通道性能劣化智能预警方法，其特征在于，所述将所述性能趋势预测结果与所述分时动态阈值比较，判断所述待监测光通道是否出现当前性能异常，包括：
    从所述性能趋势预测结果中获取当前时刻对应的当前性能数据，并从所述分时动态阈值中获取所述当前时刻的当前动态阈值；
    判断所述当前性能数据是否为无效数据；
    在所述当前数据不为无效数据时，获取所述待监测光通道的光通道性能静态阈值；
    将所述当前性能数据分别与所述光通道性能静态阈值及所述当前动态阈值的比较；
    在所述当前性能数据不大于所述光通道性能静态阈值及所述当前动态阈值时，判定所述待监测光通道未出现当前性能异常；
    在所述当前性能数据大于所述光通道性能静态阈值和/或所述当前动态阈值时，判定所述待监测光通道出现当前性能异常。
  10. 如权利要求6所述的光通道性能劣化智能预警方法，其特征在于，所述在所述待监测光通道出现性能劣化时，进行预警处理，包括：
    在所述待监测光通道出现性能趋势劣化时，获取当前性能异常和/或趋势异常的异常次数；
    根据所述异常次数从预设预警策略中确定对应的目标预警处理策略，根据所述目标预警处理策略进行预警通知。
  11. 如权利要求2-10中任一项所述的光通道性能劣化智能预警方法，其特征在于，所述获取电信传送网中若干待监测光通道的性能数据之后，所述光通道性能劣化智能预警方法还包括：
    对所述性能数据进行无效数据检查；
    在所述性能数据中出现目标无效数据时，获取所述目标无效数据的上一预设时刻或下一预设时刻的有效性能值；
    将所述目标无效数据替换为所述有效性能值，将替换后的性能数据作为新的性能数据。
  12. 一种光通道性能劣化智能预警装置，其特征在于，所述光通道性能劣化智能预警装置包括：
    模型构建模块，用于获取电信传送网中待监测光通道的性能数据，根据所述性能数据构建所述待监测光通道的关键绩效指标KPI性能预测模型；
    预测模块，用于根据KPI性能预测模型对所述待监测光通道的性能进行周期趋势预测，获得性能趋势预测结果；
    预警模块，用于根据性能趋势预测结果判断所述待监测光通道是否出现性能劣化，在出现性能劣化时，进行预警处理。
  13. 如权利要求12所述的光通道性能劣化智能预警装置，其特征在于，所述模型构建模块包括：
    性能获取模块，用于获取电信传送网中若干待监测光通道的性能数据；
    算法选取模块，用于根据各性能数据确定各待监测光通道的各性能预测目标，根据各性能预测目标从预设数据库中选取对应的若干目标模型算法；
    模型生成模块，用于根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测模型。
  14. 如权利要求13所述的光通道性能劣化智能预警装置，其特征在于，所述模型生成模块包括：
    基线模型构建模块，用于根据各性能预测目标和各目标模型算法构建各待监测光通道的关键绩效指标KPI性能预测候选基线模型；
    选取模块，用于获取各待监测光通道的历史性能数据，根据所述历史性能数据从各KPI性能预测候选基线模型中选取最优基线模型；
    参数获取模块，用于获取若干待监测光通道的通道参数；
    划分模块，用于根据各通道参数将各待监测光通道划分为相同路由不同波长的光通道组；
    分组训练模块，用于基于所述最优基线模型对各光通道组进行分组训练，获得各待监测光通道的KPI性能预测模型。
  15. 一种光通道性能劣化智能预警设备，其特征在于，所述光通道性能劣化智能预警设备包括：存储器、处理器及存储在所述存储器上并可在所述处理器上运行的光通道性能劣化智能预警程序，所述光通道性能劣化智能预警程序配置为实现如权利要求1至11中任一项所述的光通道性能劣化智能预警方法的步骤。
  16. 一种存储介质，其特征在于，所述存储介质上存储有光通道性能劣化智能预警程序，所述光通道性能劣化智能预警程序被处理器执行时实现如权利要求1至11中任一项所述的光通道性能劣化智能预警方法的步骤。
PCT/CN2020/110176 2020-06-11 2020-08-20 光通道性能劣化智能预警方法、装置、设备及存储介质 WO2021248690A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
BR112022019407A BR112022019407A2 (pt) 2020-06-11 2020-08-20 Método inteligente e aparelho de alerta antecipado para degradação de desempenho do canal óptico, dispositivo e meio de armazenamento.

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010531639.8 2020-06-11
CN202010531639.8A CN111865407B (zh) 2020-06-11 2020-06-11 光通道性能劣化智能预警方法、装置、设备及存储介质

Publications (1)

Publication Number Publication Date
WO2021248690A1 true WO2021248690A1 (zh) 2021-12-16

Family

ID=72986075

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/110176 WO2021248690A1 (zh) 2020-06-11 2020-08-20 光通道性能劣化智能预警方法、装置、设备及存储介质

Country Status (3)

Country Link
CN (1) CN111865407B (zh)
BR (1) BR112022019407A2 (zh)
WO (1) WO2021248690A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114389688A (zh) * 2022-01-04 2022-04-22 烽火通信科技股份有限公司 一种光模块性能预测的方法和装置
CN114444608A (zh) * 2022-02-08 2022-05-06 中国电信股份有限公司 数据集质量评估方法及装置、电子设备及存储介质
CN116644351A (zh) * 2023-06-13 2023-08-25 石家庄学院 一种基于人工智能的数据处理方法及系统
CN116738689A (zh) * 2023-05-26 2023-09-12 中国长江电力股份有限公司 基于Elman模型的水轮发电机组状态健康评估系统及方法
CN117390898A (zh) * 2023-12-12 2024-01-12 嘉一达智能科技有限公司 一种医疗线缆可靠性预测方法及系统

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113037365B (zh) * 2021-03-02 2022-06-24 烽火通信科技股份有限公司 一种识别光通道生命周期运维状态的方法与装置
CN116340433B (zh) * 2023-05-31 2023-07-28 中国水利水电第七工程局有限公司 施工监测信息存储计算方法、存储介质、设备及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106059661A (zh) * 2015-12-25 2016-10-26 国家电网公司 一种基于时序分析的光传输网络趋势预测方法
CN110046744A (zh) * 2019-03-12 2019-07-23 平安科技（深圳）有限公司 基于趋势预测的能耗数据预警方法及相关设备
US10396897B1 (en) * 2018-04-17 2019-08-27 General Electric Company Systems and methods for predicting defects in optical transceiver devices
CN110555546A (zh) * 2019-07-31 2019-12-10 烽火通信科技股份有限公司 一种光性能劣化趋势预测的更新方法及系统
CN111030753A (zh) * 2019-12-23 2020-04-17 国家电网有限公司信息通信分公司 基于性能与业务通道的电力通信网深度巡检方法及系统

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7716571B2 (en) * 2006-04-27 2010-05-11 Microsoft Corporation Multidimensional scorecard header definition
US10911318B2 (en) * 2015-03-24 2021-02-02 Futurewei Technologies, Inc. Future network condition predictor for network time series data utilizing a hidden Markov model for non-anomalous data and a gaussian mixture model for anomalous data
WO2018002691A1 (en) * 2016-06-29 2018-01-04 Telefonaktiebolaget Lm Ericsson (Publ) Dynamic radio downlink power amplifier control for base station energy efficiency
CN106709580B (zh) * 2017-01-13 2021-06-08 国家电网公司 一种变电站二次系统运维云平台
CN110198223B (zh) * 2018-02-27 2022-01-28 中兴通讯股份有限公司 网络故障预测方法、装置及设备、存储介质
CN110635952B (zh) * 2019-10-14 2021-03-16 中兴通讯股份有限公司 通信系统的故障根因分析方法、系统和计算机存储介质
CN111143102B (zh) * 2019-12-13 2024-01-19 东软集团股份有限公司 异常数据检测方法、装置、存储介质及电子设备
CN110995508B (zh) * 2019-12-23 2022-11-11 中国人民解放军国防科技大学 基于kpi突变的自适应无监督在线网络异常检测方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106059661A (zh) * 2015-12-25 2016-10-26 国家电网公司 一种基于时序分析的光传输网络趋势预测方法
US10396897B1 (en) * 2018-04-17 2019-08-27 General Electric Company Systems and methods for predicting defects in optical transceiver devices
CN110046744A (zh) * 2019-03-12 2019-07-23 平安科技（深圳）有限公司 基于趋势预测的能耗数据预警方法及相关设备
CN110555546A (zh) * 2019-07-31 2019-12-10 烽火通信科技股份有限公司 一种光性能劣化趋势预测的更新方法及系统
CN111030753A (zh) * 2019-12-23 2020-04-17 国家电网有限公司信息通信分公司 基于性能与业务通道的电力通信网深度巡检方法及系统

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114389688A (zh) * 2022-01-04 2022-04-22 烽火通信科技股份有限公司 一种光模块性能预测的方法和装置
CN114389688B (zh) * 2022-01-04 2023-06-13 烽火通信科技股份有限公司 一种光模块性能预测的方法和装置
CN114444608A (zh) * 2022-02-08 2022-05-06 中国电信股份有限公司 数据集质量评估方法及装置、电子设备及存储介质
CN114444608B (zh) * 2022-02-08 2024-03-29 中国电信股份有限公司 数据集质量评估方法及装置、电子设备及存储介质
CN116738689A (zh) * 2023-05-26 2023-09-12 中国长江电力股份有限公司 基于Elman模型的水轮发电机组状态健康评估系统及方法
CN116644351A (zh) * 2023-06-13 2023-08-25 石家庄学院 一种基于人工智能的数据处理方法及系统
CN116644351B (zh) * 2023-06-13 2024-04-02 石家庄学院 一种基于人工智能的数据处理方法及系统
CN117390898A (zh) * 2023-12-12 2024-01-12 嘉一达智能科技有限公司 一种医疗线缆可靠性预测方法及系统
CN117390898B (zh) * 2023-12-12 2024-03-12 嘉一达智能科技有限公司 一种医疗线缆可靠性预测方法及系统

Also Published As

Publication number Publication date
BR112022019407A2 (pt) 2023-01-31
CN111865407A (zh) 2020-10-30
CN111865407B (zh) 2021-11-30

Similar Documents

Publication Publication Date Title
WO2021248690A1 (zh) 光通道性能劣化智能预警方法、装置、设备及存储介质
US11403164B2 (en) Method and device for determining a performance indicator value for predicting anomalies in a computing infrastructure from values of performance indicators
US11409962B2 (en) System and method for automated insight curation and alerting
US20090171879A1 (en) Systems and/or methods for prediction and/or root cause analysis of events based on business activity monitoring related data
US20190163546A1 (en) Correlating failures with performance in application telemetry data
EP2466495B1 (en) Systems and/or methods for event stream deviation detection
US20050097207A1 (en) System and method of predicting future behavior of a battery of end-to-end probes to anticipate and prevent computer network performance degradation
US20170169143A1 (en) System for maintenance recommendation based on performance degradation modeling and monitoring
US11307916B2 (en) Method and device for determining an estimated time before a technical incident in a computing infrastructure from values of performance indicators
US20220334904A1 (en) Automated Incident Detection and Root Cause Analysis
CN114465874B (zh) 故障预测方法、装置、电子设备与存储介质
US11675643B2 (en) Method and device for determining a technical incident risk value in a computing infrastructure from performance indicator values
CN111290913A (zh) 一种基于运维数据预测的故障定位可视化系统和方法
US11899444B2 (en) Real-time alerts and transmission of selected signal samples under a dynamic capacity limitation
CN117391373B (zh) 一种基于多维数据的全渠道智慧运维管理方法及系统
US20240346049A1 (en) Dynamic resolution estimation for a detector
Mdini et al. Monitoring the network monitoring system: Anomaly Detection using pattern recognition
CN117041312A (zh) 基于物联网的企业级信息技术监控系统
CN115688016A (zh) 一种用于大型数据库智能运维的一体化方法
EP2882139B1 (en) System and method for IT servers anomaly detection using incident consolidation
US20240248784A1 (en) Automated Incident Detection and Root Cause Analysis
CN117573412A (zh) 系统故障预警方法、装置、电子设备及存储介质
CN116680289A (zh) 分布式数据库索引推荐监控系统及方法
CN116010897A (zh) 数据异常检测的方法、装置、电子设备及存储介质
Wang et al. Anomaly detection for mobile network management

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20939842

Country of ref document: EP

Kind code of ref document: A1

REG Reference to national code

Ref country code: BR

Ref legal event code: B01A

Ref document number: 112022019407

Country of ref document: BR

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 112022019407

Country of ref document: BR

Kind code of ref document: A2

Effective date: 20220927

122 Ep: pct application non-entry in european phase

Ref document number: 20939842

Country of ref document: EP

Kind code of ref document: A1