CN109872003A - Obj State prediction technique, system, computer equipment and storage medium - Google Patents
Obj State prediction technique, system, computer equipment and storage medium Download PDFInfo
- Publication number
- CN109872003A CN109872003A CN201910162913.6A CN201910162913A CN109872003A CN 109872003 A CN109872003 A CN 109872003A CN 201910162913 A CN201910162913 A CN 201910162913A CN 109872003 A CN109872003 A CN 109872003A
- Authority
- CN
- China
- Prior art keywords
- index
- model
- state
- distribution
- data
- Prior art date
- Legal status (The legal status 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 status listed.)
- Granted
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention discloses a kind of Obj State prediction technique, system, computer equipment and storage medium, wherein Obj State prediction technique, it include: to be grouped according to knowledge base and associated data excavation to the monitoring data index of monitoring object, and state analysis model is constructed, the state analysis model includes: single index value prediction model, main index forecast of distribution model and Obj State classification prediction model;Model training is carried out to state analysis model using sample data, obtains computation model system;The current data flow of status predication object is inputted into computation model system, Obj State is predicted.The present invention can effective lifting means O&M intelligent monitoring.
Description
Technical field
The present invention relates to big data and field of artificial intelligence more particularly to a kind of Obj State prediction technique, system,
Computer equipment and storage medium.
Background technique
With the continuous development of communications industry, gradually form comprehensive, the second best in quality communications infrastructure of covering with communicate
Network realizes wireless communication for various terminals user.In this underlying communication network, a large amount of base stations constitute its interior joint.Base
Stand by 4G based on 5G technology, provide the service such as networking, data exchange, communication for peripheral terminal equipment.The covering of one base station
Range includes multiple serving cells, and each serving cell provides communication to the terminal user within the scope of surrounding sites particular space and protects
Barrier.
Communication base station and cell, according to certain setting condition, set itself during providing radio communication service
Standby performance parameter is configured, and equipment forms service supply to the terminal user on periphery under this configuration parameter.It is transported in equipment
During row, by originals such as the fluctuation of service area number of users, communication flows variation, periphery interference and the transmitting of other cell failures
Cause will affect the performance of this cell, then telecommunication service quality be caused to decline, and shows as intra-cell users and connects into power drop
Low, user network uplink and downlink slows, and situations such as a large amount of network access failures.Therefore, operator need for equipment,
Operation security mechanism is established in cell and base station, and by lasting detection specific indexes, various exceptions, failure conditions are caught in realization
It catches and disposes, to reduce the scale of failure generation, reduce fault incidence, improve the experience of terminal user's radio communication service
Sense.
Currently, operator is in operational system build process, based on equipment mechanism and expertise, with " data
Related operation management system is constructed based on acquisition-threshold test-event recognition " mode.During actual operation, due to not
Different with cell service condition, running environment dynamic change, above-mentioned static threshold mode is unable to satisfy the dynamic under actual conditions
The status monitoring of variation, and these problems cause the working efficiency in practical maintenance work lower, in daily monitoring, because of threshold
The problem of value is arranged leads to largely false alert appearance, due to communicating the particularity of maintenance work, after the police of these falsenesses is required to
Continuous disposition, then causes entire maintenance work cost persistently to rise, influences the performance of communications infrastructure efficiency.Simultaneously as
The communications infrastructure is huge, can not specific aim give cell to establish personalized threshold monitor system, therefore these problems are to logical
Believe that the operation management of infrastructure forms larger impact, it would be highly desirable to solve.
Summary of the invention
The present invention provides a kind of Obj State prediction technique, system, computer equipment and storage medium, above-mentioned to overcome
The technical problems existing in the prior art realize equipment O&M intelligence prison to carry out health state evaluation and prediction to equipment
It surveys.
The present invention provides a kind of Obj State prediction technique, comprising:
The monitoring data index of monitoring object is grouped according to knowledge base and associated data excavation, and constructs state point
Model is analysed, the state analysis model includes: single index value prediction model, main index forecast of distribution model and Obj State point
Class prediction model;
Model training is carried out to state analysis model using sample data, obtains computation model system;
The current data flow of status predication object is inputted into computation model system, Obj State is predicted.
The present invention also provides a kind of Obj State forecasting systems, comprising:
Modeling module, for being divided according to knowledge base and associated data excavation the monitoring data index of monitoring object
Group, and state analysis model is constructed, the state analysis model includes: single index value prediction model, main index forecast of distribution mould
Type and Obj State classification prediction model;
Training module obtains computation model system for carrying out model training to state analysis model using sample data;
Prediction module, for the current data flow of status predication object to be inputted computation model system, to Obj State into
Row prediction.
The present invention also provides a kind of computer equipment, including memory and processor, being stored on the memory can be
The computer program run on the processor, the processor execute above-mentioned Obj State when running the computer program
The step of prediction technique.
The present invention also provides a kind of storage medium, the storage medium is stored with computer program, the computer program
The step of above-mentioned Obj State prediction technique can be performed.
The embodiment of the present invention is by using the single index value prediction model based on time series in building state analysis model
Reference index is predicted, and then is predicted with the associated main index for influencing Obj State of reference index, thus real
The dynamic prediction to state influence factor is showed, by building and physical training condition analysis model forms practical computation model body
System, provides solution for the Intellectualized monitoring of mass data, and therefore, of the invention not only improves Obj State prediction
Accuracy, also help Intellectualized monitoring, for Large Complex Equipment monitoring, traffic, logistics, smart city and environment protect
Shield etc. has positive application value.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without any creative labor, can be with
It obtains other drawings based on these drawings.
Fig. 1 is a kind of Obj State prediction technique flow chart provided in an embodiment of the present invention;
Fig. 2 is technical architecture plan provided in an embodiment of the present invention;
Fig. 3 is the flow chart of the mode excavation of index distribution in the embodiment of the present invention;
Fig. 4 is the flow chart of state analysis model training in the embodiment of the present invention;
Fig. 5 is the training process flow chart of single index value prediction model in the embodiment of the present invention;
Fig. 6 is the schematic network structure of main index forecast of distribution model in the embodiment of the present invention;
Fig. 7 is the training flow chart of main index forecast of distribution model in the embodiment of the present invention;
Fig. 8 is the schematic network structure of state classification prediction model in the embodiment of the present invention;
Fig. 9 is the training flow chart of main index forecast of distribution model in the embodiment of the present invention;
Figure 10 is the flow diagram predicted in the embodiment of the present invention Obj State;
Figure 11 is the specific flow chart calculated in the embodiment of the present invention Obj State prediction;
Figure 12 is a kind of structural schematic diagram of Obj State forecasting system provided in an embodiment of the present invention;
Figure 13 is the structural schematic diagram of another Obj State forecasting system provided in an embodiment of the present invention.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with attached in the embodiment of the present invention
Figure, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is the present invention
A part of the embodiment, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not having
Every other embodiment obtained under the premise of creative work is made, shall fall within the protection scope of the present invention.
To keep technical solution of the present invention clearer, the embodiment of the present invention is carried out specifically below in conjunction with attached drawing
It is bright.
The embodiment of the present invention is with O&M maintenance in magnanimity base station is application scenarios in mobile communications industry, for wherein O&M pipe
Science and engineering makees the growth requirement of intelligent fining, and the embodiment of the present invention provides the computing architecture of a state analysis and prediction.?
In this framework, Obj State is classified according to business scope knowledge first.On the basis of this, in conjunction with expertise with
And association mining, form " the main index-reference index of state-" three-level analysis model.Around this model, reference index is established
Time Series Analysis Model is realized and is predicted its value.On the basis of this, by way of Joint Distribution probability (Bayesian network)
It realizes and the Distribution value of main index is predicted.BP neural network is finally utilized, it is real on the basis of multiple main index value forecast of distribution
Now the classification of Obj State is predicted.
Fig. 1 is a kind of Obj State prediction technique flow chart provided in an embodiment of the present invention, and Fig. 2 mentions for the embodiment of the present invention
The technical architecture plan of confession, in the embodiment of the present invention process shown in FIG. 1 can technical architecture plan as shown in Figure 2 realize, the present invention
On the basis of comprehensive multi-index monitoring, a kind of technology based on time series forecasting is proposed for Condition Prediction of Equipment.The skill
Art structure is as shown in Fig. 2, entire computing architecture is constituted by modeling, training excavation and calculating to analyze three links.Wherein:
Modeling: the main target of this link is to establish Obj State classification, and establishes index grouping, in multiple indexs
Numerical value transitive relation is established between grouping, ultimately forms the connection relationship between state classification.Many levels are consequently formed
Index-state network structure.In modeling process, first on the basis of monitored object acquires data target collection, in conjunction with special
The means such as family's domain knowledge, mechanism model, filter out the specific index of a collection of representation pattern.On this basis, known according to correlation
Know or association mining means establish index grouping.In these groupings, the index of a part grouping forms directly Obj State
Value transitive relation, this grouping is known as the grouping of main index, reference index grouping is grouped into locating for other indexs.It is therein
For reference index in day-to-day operation, the transformation of value does not directly affect Obj State, but by transitive relation, influence main finger
The main index with incidence relation in mark grouping.In the present invention, by reference to referring to target value prediction and variation transmitting, main finger
Mark distribution judgement, Obj State classification three-level relational implementation prediction.
In three-level model, a main index may be relevant with one or more reference indexs.Under normal conditions,
Predictor calculation of the Distribution value of main index Jing Guo reference index obtains, and main index oneself can also independently be made under to a certain degree
For the single index value prediction processing based on time series.
Training is excavated: in this link, the training of each model is realized by bringing sample data into.First choice is according to reference
The sample time-series data of index are trained, and obtain the corresponding value prediction model of the index;Then, by reference index
Total data clustered, obtain corresponding Distribution value cluster result.In this way, predicted value above-mentioned can be mapped in some cluster
In, predicted value is subjected to sliding-model control;Then, Bayesian network will be constructed between reference index and corresponding main index and instruct
Practice, obtains main index joint probability distribution prediction model;Finally, constructing neural network on the basis of multiple main indexs, passing through
Sample training forms the state classification model based on multiple main indexs, completes to obtain after training is excavated and completes computation model system.
Calculate analysis: in this link, i.e. status predication link, access real-time stream, when the ti moment is to ti+1
The state for carving object is predicted.In this course, the corresponding data of reference index in incoming data stream are handled simultaneously
Predict it in the value at ti+1 moment.It is clustered in corresponding Distribution value cluster after obtaining this value, obtains prediction distribution
As a result.It then is updated to joint probability distribution prediction model as input using all referring to the prediction distribution result of index, is calculated
All maximum possible distribution of the main index at the ti+1 moment.Then state classification mould is updated to using this result as input
End-state prediction result is obtained in type, to realize the status predication to object.
As shown in Figure 1, the Obj State prediction technique in the present embodiment, comprising:
Step 10 is grouped the monitoring data index of monitoring object according to knowledge base and associated data excavation, and structure
Build state analysis model, the state analysis model includes: single index value prediction model, main index forecast of distribution model and right
As state classification prediction model.
In this step, the specific index of representation pattern is filtered out from the monitoring data of monitoring object concentration according to knowledge base;
Then index grouping is carried out using association mining method, index is divided into main index and reference index, wherein the main direct shadow of index
Obj State is rung, the variation of reference index is passed to main index by incidence relation by reference index.
Wherein, single index value prediction model is to carry out value prediction to reference index based on time series, and main index distribution is pre-
It surveys model and realizes that multifactor associated main index forecast of distribution, Obj State classify prediction model using mind using Bayesian network
The classification of Obj State is predicted through network implementations.In practical application, according to the scale and/or calculated performance of access analysis object
Demand, the number of plies of neural network middle layer and every layer of number of nodes elastic can configure.
Since a monitored object has multiple observation indexs, using these indexs as foundation, monitoring system realizes data
Acquisition, during status predication, the calculating for needing to rely on these indexs realizes the classification to state and forms prediction conclusion.Cause
This needs to initially set up the dependence and value TRANSFER MODEL between index-state in the present invention, this model is substantially fixed
Justice is as follows:
StateModel={ stateDef, mainParams, ReferParams }
Wherein:
StateDef is Obj State definition set, and there is defined object whole status informations, is defined as follows:
StateDef={ NaN, health, risk, alert, sick }
Wherein
NaN, which indicates that object is currently at, can not know state, be typically due to quality of data deficiency or loss data cause;
Health indicates that object current state is good, performance degradation does not occur or the case where service condition reaches boundary;
Risk indicates that existing object is in risk status, certain performance degradation occurs or part main indicator is overflowed
The case where security boundary;
Sick indicate existing object be in failure disabled state, be not normally functioning;
For the angle of fault progression rule, the deterioration sequence of above-mentioned state is expressed as
Dis (health, sick) < dis (risk, sick)
MainParams is main index set in a model.It include one group of observation index, these indexs in main index set
The exception of numerical value directly cause object some properties failure disability generation, be defined as follows:
MainParams={ (A, w, vCluster, CPs, CAs)i| i=1,2 ... .n }
It is made of an index group, in index group:
A is expressed as the mark of the main index;
W is weighing factor of the main index to end-state, this weighted value needs to obtain by model training;
VCluster is that the codomain of the main index is distributed gathering conjunction, is made of one group of cluster information, this distribution gathering is closed
It needs to obtain by excavating;
CPs is the main index and the corresponding reference index joint probability distribution table with incidence relation.It is defined as follows:
CPs={ (P (A ∈ vi|∑Ck∈Cvp))j| j=1,2 ... .m }
vi∈vCluster
Ck∈ ReferParams, Cvp∈Ck
CAs is transitive relation net, and there is defined the incidence relations between current reference index and main index, and definition is such as
Under:
CAs={ CP| p=1,2 ... n }
Cp∈ReferParams
Expertise is crossed by this transitive relation Netcom or index incidence relation excavates, such as: Apriori method realizes structure
It builds, specific method repeats no more.
ReferParams is reference index collection, wherein containing one group of reference index, the data of these indexs are not direct
React the variation degree of object property.But fluctuation or distribution of the value in time series will affect master by transitive relation
The Distribution value of index then leads to the transformation of Obj State.It is defined as follows:
ReferParams={ (C, CAs)i| i=1,2 ... .m }
C is reference index, is defined as follows:
C={ Ckey, { Cvp| p=1,2 ... n } }
Wherein, Ckey is the title of the reference index;
CvpIt is closed for the Distribution value gathering of the index, this cluster is obtained by excavating.
By the above method, " the main index-reference index of state-" three-level model is constructed, surrounds this model, firstly,
To reference index settling time series analysis model, realizes and its value is predicted;Secondly, passing through Joint Distribution probability (Bayesian network)
Mode realize to the prediction of the Distribution value of main index;BP neural network is finally utilized, in the base of multiple main index value forecast of distribution
On plinth, realize that the classification to Obj State is predicted.
It is above-mentioned construct three-level model after, need to excavate index value distribution pattern in order to carry out prediction meter
It calculates.Fig. 3 is the flow chart of the mode excavation of index distribution in the embodiment of the present invention, and mode excavation is referred mainly to for index Distribution value
Mode is excavated, and according to daily monitoring data, is carried out clustering processing to data distribution, is formed distribution cluster, then realize to even
The discretization of continuous Wave data.
The data of index are clustered, distribution cluster is formed and are marked.In the analysis process, by sampled data or in advance
Measured value is calculated with clustering cluster, obtains the distribution cluster information of sampled data or predicted value;It is real by the extraction for being distributed cluster information
The sliding-model control of existing sampled data and predicted value improves model computational efficiency as the input of prediction model.
In the present invention, mode excavation is related to main index and reference index, and the Distribution value mode of the main index of object is stored in
In aforementioned vCluster, the Distribution value mode of reference index is stored in Cvp.These two types of index value distribution patterns are excavated using same
The algorithm of sample, concrete processing procedure are as shown in Figure 3.
Step 20 carries out model training to state analysis model using sample data, obtains computation model system;
Fig. 4 is the flow chart of state analysis model training in the embodiment of the present invention, and state analysis model is main in the present invention
Including single index value prediction model, multifactor associated main index forecast of distribution model and Obj State based on time series
Classification prediction model.The training process of these models is specifically as shown in Figure 4:
Step 201 selects smoothing processing function according to the corresponding sample data characteristic distributions of reference index, by initial data
It is converted to data sequence and brings the single index value prediction model of building into and be trained iteration, obtain stable based on time series
Single index value prediction model;
Step 202, the main finger of expression for bringing the distribution cluster information of the sample data of reference index into building as trained values
It marks and is trained with the Bayesian network of reference index incidence relation, obtain stable multifactor associated main index forecast of distribution
Model;
Step 203 brings the distribution cluster information of main index predicted value into building predict Obj State as trained values
It is trained in the disaggregated model of classification, obtains stable Obj State classification prediction model.
Above-mentioned model and model training process difference are as follows:
1, single index value prediction model and model training based on time series
Fig. 5 is the training process flow chart of single index value prediction model in the embodiment of the present invention, as shown in figure 5, this mould
Type is mainly for reference index, on the basis of time series, establishes prediction model, in the present invention algorithm needed for prediction model
Using Arima algorithm.In the training process, suitable smoothing method is selected according to the characteristics of reference index numerical transformation first,
Including the methods of D order difference, logistics, data sequence is formed after initial data is converted, then by the data sequence
It brings into model, after successive ignition, forms the stable single index value prediction model based on time series.
2, multifactor associated main index forecast of distribution model and model training
Fig. 6 is the schematic network structure of main index forecast of distribution model in the embodiment of the present invention, as shown in fig. 6, this
Model is to construct Bayesian network on the basis of main index and reference index incidence relation, based on the corresponding application scenarios of the present invention
The characteristics of middle object scale, improves efficiency to reduce calculation amount, and Bayesian network is using simple layer network structure as schemed
Shown in 6, this model predicts that Bayesian network, this network are defeated by an input layer and one for each main Index Establishment
Layer is constituted out.Each node is the reference index for having incidence relation with the main index in input layer, and output layer is the main index
Corresponding Joint Distribution probability tables.Connection pass is established between reference index C1, C2 ... the ..Cn node and output layer of input layer
It is that the probability tables between the input node and output node are recorded in connection relationship.
Fig. 7 is the training flow chart of main index forecast of distribution model in the embodiment of the present invention, as shown in fig. 7, in model structure
After building, bring sample data into and be trained, in the training process, first choice is needed C1, the sample data of C2 ... ..Cn into
The processing of row value, obtains its corresponding distribution cluster information, brings network into using distribution cluster information as trained values, is formed surely by iteration
Fixed Bayesian network, detailed process are as shown in Figure 7.
3, Obj State classification prediction model and model training
After obtaining the distribution of forecasting value of multiple main indexs, so that it may rely on shape of the predicted value to object of these main indexs
State is classified, and then realizes prediction.As previously mentioned, in the present invention by the state of object be divided into NaN, health, risk,
The several ranks of sick, the discrete expression of status information is realized by this mode.
Fig. 8 is the schematic network structure of state classification prediction model in the embodiment of the present invention, as shown in figure 8, in this hair
In bright, the characteristics of for application scenarios, realize that the joint classification of multiple main indexs is predicted by BP neural network.In conjunction with the present invention
The larger feature of object in corresponding application scenarios, this neural network use telescopic hierarchical relationship.For right
As status predication establishes BP neural network, using main index distribution of forecasting value cluster above-mentioned as network inputs section in this network
Point, each node corresponds to some main index in input layer, and middle layer is scalable layer, is required according to overall computational performance, right
The number of plies and every layer of number of nodes carry out elastic telescopic configuration.Effect is calculated in application scenarios mentioned by the present invention in order to improve
Rate, using one layer of structure, the hidden unit number in this layer is consistent with the node number of input layer, and output layer is corresponding state
Classification results.
Fig. 9 is the training flow chart of main index forecast of distribution model in the embodiment of the present invention, as shown in figure 9, in model structure
After building, bring sample data into and be trained, in the training process, first choice is needed A1, the sample data of A2 ... ..Am into
The processing of row value, obtains its corresponding Distribution value cluster information, network is brought into using Distribution value cluster information as trained values, by iteration shape
At stable sorter network.
In practical applications, for the characteristics of object and the requirement of calculated performance, can be used not in different application scene
Same sorting algorithm adaptation, realizes flexible adaptation.
The current data flow of status predication object is inputted computation model system by step 30, is predicted Obj State.
Figure 10 is the flow diagram predicted in the embodiment of the present invention Obj State, is such as schemed in virtual condition prediction
Shown in 10, include the following steps:
The current data flow of monitored object is sampled and is clustered by step 301, obtains the monitored object currently main index
Distribution cluster information;
The distribution cluster information of current main index is brought into computation model system by step 302, and detection monitored object is presently in
State the sampled data of the corresponding reference index of the main index of existing object is then obtained, to sampled data if abnormal condition
Time series forecasting is carried out, and obtains the distribution cluster information of corresponding reference index according to predicted value;
The distribution cluster information of reference index is brought into the main index forecast of distribution mould after corresponding training by step 303 respectively
Type obtains the distribution cluster information of main index;
Step 304, prediction model that the Obj State that the distribution cluster information of main index is brought into after training is classified, obtain object
Predicted state.
In the monitoring activity for mass object, the present invention is real to different Obj States from algorithm using model above-mentioned
Existing intellectual analysis.In this course, application system constructs sample set first with historical data, on the basis of sample set,
For each object, the training of aforementioned correlation distribution pattern excavation and prediction model is completed, complete analysis knowledge is formed
Library.On this basis, it is analyzed for incremental data, realizes the identification and early warning of Obj State.
Figure 11 is the specific flow chart calculated in the embodiment of the present invention Obj State prediction, as shown in figure 11, in early warning
In the process, first to main index value carry out distribution cluster detection, if object main index distribution cluster testing result do not occur it is different
Reason condition, and it is also abnormal without occurring after status switch detects, then current object state is labeled as normal condition
health;If it is abnormal that part occurs in the main index distribution cluster testing result of existing object, its reference index data is extracted, is brought into
Status predication is carried out in aforementioned prediction model, according to the status information of result queue existing object, is occurred as Obj State has
Obj State is then labeled as prediction result risk by the possibility changed extremely, if Obj State does not have the possibility for being abnormal,
Then mark normal condition health;If larger range exception occurs in the main index distribution cluster testing result of existing object, mark
Current state sick simultaneously carries out succeeding state prediction.
In above-mentioned status predication labeling process, to improve working efficiency, it can increase before predicted state to object
Whether state, which reaches Anomaly standard, is judged, i.e., when the state for detecting that monitored object is presently in is abnormal condition,
And Obj State reaches preset Anomaly standard, then can no longer need to carry out prediction calculating with Direct Mark Obj State to judge
With the calculation amount for the system of saving and status predication time;And works as and detect that the state that monitored object is presently in is abnormal condition
When, and it is pre- further in accordance with the prediction technique that above-mentioned prediction model uses to carry out state when Obj State is not up to preset Anomaly standard
It surveys, preset Anomaly standard can occur abnormal for the main index in specific part, or multiple main indexs occur different simultaneously
Often.
In embodiments of the present invention, in the real-time status detection process of mass object, mould is carried out to main index first
Formula matching, identifies the state of existing object, and there is improper object just to enter subsequent analytical calculation process for identified confirmation;
Obj State be identified as it is improper after, according to the pattern match result tagged object current state of main index, according to aforementioned three
Grade the model calculation obtains its predicted state, and result is labeled and is issued.
The embodiment of the present invention is by using the single index value prediction model based on time series in building state analysis model
Reference index is predicted, and then is predicted with the associated main index for influencing Obj State of reference index, thus real
The dynamic prediction to state influence factor is showed, by building and physical training condition analysis model forms practical computation model body
System, provides solution for the Intellectualized monitoring of mass data, and therefore, of the invention not only improves Obj State prediction
Accuracy, also help Intellectualized monitoring, for Large Complex Equipment monitoring, traffic, logistics, smart city and environment protect
Shield etc. has positive application value.
Problem alert for the falseness occurred in operational system above-mentioned, base of the present invention in big data and artificial intelligence technology
A kind of Obj State Predicting Technique based on multi objective time series is proposed on plinth.The technology is screened on the basis of monitoring data collection
One group of index, this group of index by main index and and reference index with relevance constitute, the health status of cell is by leading
Index combined calculation obtains, and the numerical value change of reference index influences the Distribution value of main index, in calculating process, refers to for reference
Settling time sequence prediction function is marked, predicts distributed area corresponding to its value of next timing node.It on this basis, will be pre-
Input of the measured value distributed area as main index prediction model is then realized and is predicted main finger target value, final to realize to equipment
State is predicted false alert situation occur to effectively reduce in practice.
Method proposed by the invention realizes health state evaluation and prediction for special object, hands in power grid, track
It is equally applicable in the similar equipment Intellectualized monitorings such as logical, aircraft and operation management scene.
Figure 12 is a kind of structural schematic diagram of Obj State forecasting system provided in an embodiment of the present invention, as shown in figure 12,
System of the invention, comprising: modeling module 100, training module 200 and prediction module 300, wherein modeling module 100 is used for
The monitoring data index of monitoring object is grouped according to knowledge base and associated data excavation, and constructs state analysis model,
The state analysis model includes: single index value prediction model, main index forecast of distribution model and Obj State classification prediction
Model;Training module 200 obtains computation model system for carrying out model training to state analysis model using sample data;
Prediction module 300 carries out Obj State pre- for the current data flow of status predication object to be inputted computation model system
It surveys.
The embodiment of the present invention two is able to carry out the scheme of above method embodiment one, working principle and the technology reached effect
Fruit seemingly, repeats no more.
On the basis of above-mentioned Figure 12, system of the invention still further comprises cluster module, for data clusters
Discretization is convenient for computer disposal.
Figure 13 is the structural schematic diagram of another Obj State forecasting system provided in an embodiment of the present invention, such as Figure 13 institute
Show, the system of the present embodiment is on the basis of above-described embodiment two further include: cluster module 400, for by the prison of each index
It surveys sampled data or predicted value is clustered, form different distribution clusters;And to distribution cluster serializing processing, obtain discretization
Sampled data or predicted value distribution cluster information.Are inputted by each model and is analyzed for sampled value or predicted value cluster serializing
Prediction calculates.
In the specific embodiment of Obj State forecasting system of the present invention, modeling module 100 is particularly used according to knowledge base
The specific index of representation pattern is filtered out from the monitoring data of monitoring object concentration;And index point is carried out using association mining method
Index is divided into main index and reference index by group.Training module 200 is specifically used for according to the corresponding sample data point of reference index
Boot point selects smoothing processing function, and initial data is converted to the single index value prediction model progress that data sequence brings building into
Training iteration, obtains the stable single index value prediction model based on time series;By the distribution of the sample data of reference index
Cluster information brings the main index of expression of building into as trained values and the Bayesian network of reference index incidence relation is trained, and obtains
Take stable multifactor associated main index forecast of distribution model;Using the distribution cluster information of main index predicted value as trained values band
Enter and be trained in the disaggregated model to Obj State prediction classification of building, obtains stable Obj State classification prediction mould
Type.Prediction module 300 is specifically used for that the current data flow of monitored object is sampled and clustered, and it is current to obtain the monitored object
The distribution cluster information of main index;Bring the distribution cluster information of current main index into computation model system, detection monitored object is current
State in which then obtains the sampled data of the corresponding reference index of the main index of existing object, to sampling if abnormal condition
Data carry out time series forecasting, and the distribution cluster information of corresponding reference index is obtained according to predicted value;By reference index
Distribution cluster information brings the main index forecast of distribution model after corresponding training into respectively, obtains the distribution cluster information of main index;It will
The distribution cluster information of main index brings the Obj State classification prediction model after training into, obtains the predicted state of object.
The embodiment of the present invention also provides a kind of computer equipment, including memory and processor, stores on the memory
There is the computer program that can be run on the processor, the processor executes above-mentioned Fig. 1 when running the computer program
Shown in Obj State prediction technique the step of.
The embodiment of the present invention also provides a kind of storage medium, the readable storage medium storing program for executing such as: ROM/RAM, magnetic disk, CD,
Storage medium is stored with computer program, and the computer program can be by hardware devices such as terminal device, computer or servers
The step of executing above-mentioned Obj State prediction technique.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features;
And these are modified or replaceed, the range for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution.
Claims (14)
1. a kind of Obj State prediction technique characterized by comprising
The monitoring data index of monitoring object is grouped according to knowledge base and associated data excavation, and constructs state analysis mould
Type, the state analysis model include: that single index value prediction model, main index forecast of distribution model and Obj State classification are pre-
Survey model;
Model training is carried out to state analysis model using sample data, obtains computation model system;
The current data flow of status predication object is inputted into computation model system, Obj State is predicted.
2. the method according to claim 1, wherein being excavated according to knowledge base and associated data to monitoring object
Monitoring data index is grouped, specifically:
The specific index of representation pattern is filtered out from the monitoring data of monitoring object concentration according to knowledge base;
Index grouping is carried out using association mining method, index is divided into main index and reference index, wherein the main direct shadow of index
Obj State is rung, the variation of reference index is passed to main index by incidence relation by reference index.
3. the method according to claim 1, wherein further include:
The monitoring sampled data or predicted value of each index are clustered, different distribution clusters is formed;
To distribution cluster serializing processing, the sampled data of discretization or the distribution cluster information of predicted value are obtained.
4. according to the method described in claim 3, it is characterized in that, carrying out model instruction to state analysis model using sample data
Practice, obtain computation model system, specifically:
Smoothing processing function is selected according to the corresponding sample data characteristic distributions of reference index, initial data is converted to data sequence
The single index value prediction model that column bring building into is trained iteration, obtains the stable single index value prediction based on time series
Model;
The main index of expression and reference index for bringing the distribution cluster information of the sample data of reference index into building as trained values
The Bayesian network of incidence relation is trained, and obtains stable multifactor associated main index forecast of distribution model;
Bring the classification mould to Obj State prediction classification of building into using the distribution cluster information of main index predicted value as trained values
It is trained in type, obtains stable Obj State classification prediction model.
5. referring to the method according to claim 1, wherein single index value prediction model is based on time series to reference
Mark carries out value prediction.
6. the method according to claim 1, wherein main index forecast of distribution model is realized using Bayesian network
Multifactor associated main index forecast of distribution.
7. the method according to claim 1, wherein Obj State classification prediction model uses neural fusion
Classification prediction to Obj State.
8. the method according to the description of claim 7 is characterized in that according to the scale and/or calculated performance of access analysis object
Demand, the number of plies of neural network middle layer and every layer of number of nodes elastic can configure.
9. method described in claim according to claim 1~any one of 8, which is characterized in that work as status predication object
Preceding data flow inputs computation model system, predicts Obj State, specifically:
The current data flow of monitored object is sampled and is clustered, obtain the monitored object currently main index distribution cluster letter
Breath;
It brings the distribution cluster information of current main index into computation model system, detects the state that monitored object is presently in, if
Abnormal condition then obtains the sampled data of the corresponding reference index of the main index of existing object, carries out time sequence to sampled data
Column are predicted, and the distribution cluster information of corresponding reference index is obtained according to predicted value;
It brings the distribution cluster information of reference index into main index forecast of distribution model after corresponding training respectively, obtains main index
Distribution cluster information;
It brings the distribution cluster information of main index into Obj State classification prediction model after training, obtains the predicted state of object.
10. according to the method described in claim 9, it is characterized in that, when detecting that the state that monitored object is presently in is non-
When normal condition, and Obj State reaches preset Anomaly standard, then Direct Mark Obj State.
11. a kind of Obj State forecasting system characterized by comprising
Modeling module, for being grouped according to knowledge base and associated data excavation to the monitoring data index of monitoring object, and
Construct state analysis model, the state analysis model include: single index value prediction model, main index forecast of distribution model and
Obj State classification prediction model;
Training module obtains computation model system for carrying out model training to state analysis model using sample data;
Prediction module carries out Obj State pre- for the current data flow of status predication object to be inputted computation model system
It surveys.
12. system according to claim 11, which is characterized in that further include:
Cluster module forms different distribution clusters for clustering the monitoring sampled data or predicted value of each index;With
And to distribution cluster serializing processing, obtain the sampled data of discretization or the distribution cluster information of predicted value.
13. a kind of computer equipment, including memory and processor, it is stored with and can transports on the processor on the memory
Capable computer program, which is characterized in that perform claim requires to appoint in 1 to 10 when the processor runs the computer program
The step of Obj State prediction technique described in one claim.
14. a kind of storage medium, which is characterized in that the storage medium is stored with computer program, and the computer program can
The step of Obj State prediction technique being performed as described in any one of claims 1 to 10 claim.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910162913.6A CN109872003B (en) | 2019-03-06 | 2019-03-06 | Object state prediction method, object state prediction system, computer device, and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910162913.6A CN109872003B (en) | 2019-03-06 | 2019-03-06 | Object state prediction method, object state prediction system, computer device, and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109872003A true CN109872003A (en) | 2019-06-11 |
CN109872003B CN109872003B (en) | 2021-08-13 |
Family
ID=66919765
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910162913.6A Active CN109872003B (en) | 2019-03-06 | 2019-03-06 | Object state prediction method, object state prediction system, computer device, and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109872003B (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110413703A (en) * | 2019-06-21 | 2019-11-05 | 平安科技(深圳)有限公司 | The classification method and relevant device of monitor control index data based on artificial intelligence |
CN110531988A (en) * | 2019-08-06 | 2019-12-03 | 新华三大数据技术有限公司 | The trend prediction method and relevant apparatus of application program |
CN111369033A (en) * | 2020-01-02 | 2020-07-03 | 东软集团股份有限公司 | Prediction method and device for value distribution of operation and maintenance indexes |
CN111522705A (en) * | 2020-03-23 | 2020-08-11 | 广东工业大学 | Intelligent operation and maintenance solution method for industrial big data |
CN111582516A (en) * | 2020-04-29 | 2020-08-25 | 青岛聚好联科技有限公司 | Equipment maintenance method and device based on big data analysis |
CN111985713A (en) * | 2020-08-19 | 2020-11-24 | 中国银行股份有限公司 | Data index waveform prediction method and device |
CN112052914A (en) * | 2020-09-29 | 2020-12-08 | 中国银行股份有限公司 | Classification model prediction method and device |
CN112084166A (en) * | 2019-06-13 | 2020-12-15 | 上海杰之能软件科技有限公司 | Sample data establishment method, data model training method, device and terminal |
CN112785441A (en) * | 2020-04-20 | 2021-05-11 | 招商证券股份有限公司 | Data processing method and device, terminal equipment and storage medium |
CN114338416A (en) * | 2020-09-29 | 2022-04-12 | ***通信有限公司研究院 | Space-time multi-index prediction method and device and storage medium |
CN116541251A (en) * | 2023-07-04 | 2023-08-04 | 天津通信广播集团有限公司 | Display device state early warning method, device, equipment and computer readable medium |
Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2012014540A (en) * | 2010-07-02 | 2012-01-19 | Mitsubishi Electric Corp | Energy saving control device and energy saving control program |
CN103745273A (en) * | 2014-01-06 | 2014-04-23 | 北京化工大学 | Semiconductor fabrication process multi-performance prediction method |
CN104794546A (en) * | 2015-04-29 | 2015-07-22 | 武汉大学 | Wind power climbing forecasting method based on deep confidence network classifying method |
CN105225020A (en) * | 2015-11-11 | 2016-01-06 | 国家电网公司 | A kind of running status Forecasting Methodology based on BP neural network algorithm and system |
CN105468917A (en) * | 2015-12-01 | 2016-04-06 | 北京无线电计量测试研究所 | Pipeline fault prediction method and apparatus |
CN105760970A (en) * | 2016-03-21 | 2016-07-13 | 重庆灵狐科技股份有限公司 | Method for predicting AQI |
CN106447202A (en) * | 2016-09-29 | 2017-02-22 | 国网山东省电力公司 | Power equipment state evaluation method based on data mining and principal component analysis |
CN107274011A (en) * | 2017-06-05 | 2017-10-20 | 上海电力学院 | The equipment state recognition methods of comprehensive Markov model and probability net |
CN107991870A (en) * | 2017-12-05 | 2018-05-04 | 暨南大学 | A kind of fault pre-alarming and life-span prediction method of Escalator equipment |
CN108020781A (en) * | 2017-12-19 | 2018-05-11 | 上海电机学院 | A kind of circuit breaker failure diagnostic method |
CN109325811A (en) * | 2018-10-31 | 2019-02-12 | 平安直通咨询有限公司 | Value of house prediction technique, device, computer equipment and storage medium |
CN109345035A (en) * | 2018-10-31 | 2019-02-15 | 平安直通咨询有限公司 | Value of house prediction technique, device, computer equipment and storage medium |
CN109409963A (en) * | 2018-11-12 | 2019-03-01 | 平安科技(深圳)有限公司 | Prediction technique and device, storage medium, the computer equipment of customer life cycle |
-
2019
- 2019-03-06 CN CN201910162913.6A patent/CN109872003B/en active Active
Patent Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2012014540A (en) * | 2010-07-02 | 2012-01-19 | Mitsubishi Electric Corp | Energy saving control device and energy saving control program |
CN103745273A (en) * | 2014-01-06 | 2014-04-23 | 北京化工大学 | Semiconductor fabrication process multi-performance prediction method |
CN104794546A (en) * | 2015-04-29 | 2015-07-22 | 武汉大学 | Wind power climbing forecasting method based on deep confidence network classifying method |
CN105225020A (en) * | 2015-11-11 | 2016-01-06 | 国家电网公司 | A kind of running status Forecasting Methodology based on BP neural network algorithm and system |
CN105468917A (en) * | 2015-12-01 | 2016-04-06 | 北京无线电计量测试研究所 | Pipeline fault prediction method and apparatus |
CN105760970A (en) * | 2016-03-21 | 2016-07-13 | 重庆灵狐科技股份有限公司 | Method for predicting AQI |
CN106447202A (en) * | 2016-09-29 | 2017-02-22 | 国网山东省电力公司 | Power equipment state evaluation method based on data mining and principal component analysis |
CN107274011A (en) * | 2017-06-05 | 2017-10-20 | 上海电力学院 | The equipment state recognition methods of comprehensive Markov model and probability net |
CN107991870A (en) * | 2017-12-05 | 2018-05-04 | 暨南大学 | A kind of fault pre-alarming and life-span prediction method of Escalator equipment |
CN108020781A (en) * | 2017-12-19 | 2018-05-11 | 上海电机学院 | A kind of circuit breaker failure diagnostic method |
CN109325811A (en) * | 2018-10-31 | 2019-02-12 | 平安直通咨询有限公司 | Value of house prediction technique, device, computer equipment and storage medium |
CN109345035A (en) * | 2018-10-31 | 2019-02-15 | 平安直通咨询有限公司 | Value of house prediction technique, device, computer equipment and storage medium |
CN109409963A (en) * | 2018-11-12 | 2019-03-01 | 平安科技(深圳)有限公司 | Prediction technique and device, storage medium, the computer equipment of customer life cycle |
Non-Patent Citations (1)
Title |
---|
盛春阳等: "基于贝叶斯网络模型的交通状态预测", 《公路与汽运》 * |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112084166A (en) * | 2019-06-13 | 2020-12-15 | 上海杰之能软件科技有限公司 | Sample data establishment method, data model training method, device and terminal |
CN110413703B (en) * | 2019-06-21 | 2023-07-25 | 平安科技(深圳)有限公司 | Method for classifying monitoring index data based on artificial intelligence and related equipment |
CN110413703A (en) * | 2019-06-21 | 2019-11-05 | 平安科技(深圳)有限公司 | The classification method and relevant device of monitor control index data based on artificial intelligence |
CN110531988A (en) * | 2019-08-06 | 2019-12-03 | 新华三大数据技术有限公司 | The trend prediction method and relevant apparatus of application program |
CN111369033A (en) * | 2020-01-02 | 2020-07-03 | 东软集团股份有限公司 | Prediction method and device for value distribution of operation and maintenance indexes |
CN111369033B (en) * | 2020-01-02 | 2024-03-26 | 东软集团股份有限公司 | Method and device for predicting value distribution of operation and maintenance indexes |
CN111522705A (en) * | 2020-03-23 | 2020-08-11 | 广东工业大学 | Intelligent operation and maintenance solution method for industrial big data |
CN112785441B (en) * | 2020-04-20 | 2023-12-05 | 招商证券股份有限公司 | Data processing method, device, terminal equipment and storage medium |
CN112785441A (en) * | 2020-04-20 | 2021-05-11 | 招商证券股份有限公司 | Data processing method and device, terminal equipment and storage medium |
CN111582516A (en) * | 2020-04-29 | 2020-08-25 | 青岛聚好联科技有限公司 | Equipment maintenance method and device based on big data analysis |
CN111985713B (en) * | 2020-08-19 | 2023-08-18 | 中国银行股份有限公司 | Data index waveform prediction method and device |
CN111985713A (en) * | 2020-08-19 | 2020-11-24 | 中国银行股份有限公司 | Data index waveform prediction method and device |
CN114338416B (en) * | 2020-09-29 | 2023-04-07 | ***通信有限公司研究院 | Space-time multi-index prediction method and device and storage medium |
CN114338416A (en) * | 2020-09-29 | 2022-04-12 | ***通信有限公司研究院 | Space-time multi-index prediction method and device and storage medium |
CN112052914B (en) * | 2020-09-29 | 2023-12-01 | 中国银行股份有限公司 | Classification model prediction method and device |
CN112052914A (en) * | 2020-09-29 | 2020-12-08 | 中国银行股份有限公司 | Classification model prediction method and device |
CN116541251A (en) * | 2023-07-04 | 2023-08-04 | 天津通信广播集团有限公司 | Display device state early warning method, device, equipment and computer readable medium |
CN116541251B (en) * | 2023-07-04 | 2023-10-20 | 天津通信广播集团有限公司 | Display device state early warning method, device, equipment and computer readable medium |
Also Published As
Publication number | Publication date |
---|---|
CN109872003B (en) | 2021-08-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109872003A (en) | Obj State prediction technique, system, computer equipment and storage medium | |
CN109492857A (en) | A kind of distribution network failure risk class prediction technique and device | |
CN105550714A (en) | Cluster fusion method for warning information in heterogeneous network environment | |
CN105512448A (en) | Power distribution network health index assessment method | |
CN116614177B (en) | Optical fiber state multidimensional parameter monitoring system | |
CN112949874B (en) | Power distribution terminal defect characteristic self-diagnosis method and system | |
CN115423009A (en) | Cloud edge coordination-oriented power equipment fault identification method and system | |
CN111586728B (en) | Small sample characteristic-oriented heterogeneous wireless network fault detection and diagnosis method | |
CN116205265A (en) | Power grid fault diagnosis method and device based on deep neural network | |
CN111143835B (en) | Non-invasive protection method for business logic of electric power metering system based on machine learning | |
CN115964503B (en) | Safety risk prediction method and system based on community equipment facilities | |
Wang et al. | LSTM-based alarm prediction in the mobile communication network | |
CN109359810A (en) | A kind of electric power communication transmission network operating status appraisal procedure based on more strategy equilibriums | |
CN112801815B (en) | Power communication network fault early warning method based on federal learning | |
CN113807462A (en) | AI-based network equipment fault reason positioning method and system | |
CN117216722B (en) | Sensor time sequence data-based multi-source heterogeneous data fusion system | |
CN117424791B (en) | Large-scale power communication network fault diagnosis system | |
Bai et al. | Abnormal Detection Scheme of Substation Equipment based on Intelligent Fusion Terminal | |
Lu et al. | Anomaly Recognition Method for Massive Data of Power Internet of Things Based on Bayesian Belief Network | |
Zhao et al. | Research on machine learning-based correlation analysis method for power equipment alarms | |
CN112116160B (en) | Important power transmission channel disaster monitoring method based on improved cellular automaton of optimized neural network | |
CN118013470B (en) | Comprehensive input analysis method and system for intelligent environment environmental protection monitoring data | |
Meng et al. | Computer Network Security Evaluation Method Based on GABP Model | |
CN118278914A (en) | Method for realizing equipment fault rush repair based on GIS (geographic information System) | |
CN118283663A (en) | Base station fault diagnosis method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |