CN109872003B - Object state prediction method, object state prediction system, computer device, and storage medium - Google Patents
Object state prediction method, object state prediction system, computer device, and storage medium Download PDFInfo
- Publication number
- CN109872003B CN109872003B CN201910162913.6A CN201910162913A CN109872003B CN 109872003 B CN109872003 B CN 109872003B CN 201910162913 A CN201910162913 A CN 201910162913A CN 109872003 B CN109872003 B CN 109872003B
- Authority
- CN
- China
- Prior art keywords
- state
- prediction
- index
- indexes
- distribution
- 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.)
- Active
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses an object state prediction method, a system, a computer device and a storage medium, wherein the object state prediction method comprises the following steps: grouping monitoring data indexes of the monitored object according to a knowledge base and associated data mining, and constructing a state analysis model, wherein the state analysis model comprises the following steps: the system comprises a single index value prediction model, a main index distribution prediction model and an object state classification prediction model; performing model training on the state analysis model by using the sample data to obtain a calculation model system; and inputting the current data stream of the state prediction object into a calculation model system to predict the state of the object. The invention can effectively improve the intelligent monitoring of the operation and maintenance of the equipment.
Description
Technical Field
The invention relates to the technical field of big data and artificial intelligence, in particular to an object state prediction method, an object state prediction system, computer equipment and a storage medium.
Background
With the continuous development of the communication industry, communication infrastructure and communication network with comprehensive coverage and good quality are gradually formed, and wireless communication is realized for various terminal users. In this basic communication network, a large number of base stations form nodes therein. The base station provides services such as network access, data exchange, communication and the like for peripheral terminal equipment on the basis of a 4G \5G technology. The coverage area of a base station comprises a plurality of service cells, and each service cell provides communication guarantee for terminal users in a specific spatial range around the base station.
In the process of providing wireless communication service, the communication base station and the cell set performance parameters of the device according to certain setting conditions, and the device forms service supply for peripheral terminal users under the configuration parameters. In the operation process of the equipment, the performance of the cell is affected by the fluctuation of the number of users in the service area, the change of communication flow, peripheral interference, fault transmission of other cells and other reasons, so that the communication service quality is reduced, and the conditions that the connection success rate of the users in the cell is reduced, the uplink and downlink speeds of a user network are slowed down, a large amount of network access fails and the like are presented. Therefore, operators need to establish an operation support mechanism for devices, cells and base stations, and capture and handle various abnormal and failure conditions by continuously detecting specific indexes, so as to reduce the scale of occurrence of faults, reduce the fault influence range, and improve the experience of wireless communication services of terminal users.
At present, in the process of building an operation and maintenance system, an operator constructs a related operation and maintenance management system based on a data acquisition-threshold detection-event recognition mode based on an equipment mechanism and expert experience. In the actual operation process, because different cell operating conditions are different, the operating environment changes dynamically, the static threshold mode cannot meet the state monitoring of dynamic change under the actual condition, and the problems cause lower working efficiency in the actual operation and maintenance work. Meanwhile, because the communication infrastructure is huge in scale, an individualized threshold monitoring system cannot be established for a cell in a targeted manner, so that the problems have a great influence on operation and maintenance management of the communication infrastructure, and a solution is urgently needed.
Disclosure of Invention
The invention provides an object state prediction method, an object state prediction system, computer equipment and a storage medium, which are used for overcoming the technical problems in the prior art, so that health state evaluation and prediction are carried out on the equipment, and intelligent monitoring of operation and maintenance of the equipment is realized.
The invention provides an object state prediction method, which is suitable for an intelligent monitoring system for equipment operation and maintenance, and comprises the following steps:
grouping monitoring data indexes of a monitoring object according to a knowledge base and associated data mining, dividing the indexes into main indexes and reference indexes, wherein the main indexes directly influence the state of the object, the reference indexes transmit the change of the reference indexes to the main indexes through an association relation, and a state analysis model of index-state dependency relation and value transmission is constructed, and the state analysis model comprises: the method comprises the steps of carrying out value prediction on a reference index based on a time sequence through a single index value prediction model, a main index distribution prediction model for realizing multi-factor associated main index distribution prediction by adopting a Bayesian network, and an object state classification prediction model;
carrying out data acquisition according to observation indexes, clustering monitoring sampling data or predicted values of all indexes to form different distribution clusters, and carrying out serialization processing on the distribution clusters to obtain distribution cluster information of discretized sampling data or predicted values;
selecting a smooth processing function according to sample data distribution characteristics corresponding to the reference index, converting original data into a data sequence, bringing the data sequence into the constructed single-index-value prediction model, and performing training iteration to obtain a stable single-index-value prediction model based on a time sequence;
taking the distribution cluster information of the sample data of the reference index as a training value to be brought into a constructed Bayesian network representing the incidence relation between the main index and the reference index for training, and obtaining a stable multi-factor correlated main index distribution prediction model;
taking the distribution cluster information of the main index prediction value as a training value to be brought into a constructed classification model for predicting and classifying object states for training to obtain a stable object state classification prediction model;
a calculation model system is formed by the stable single index value prediction model, the main index distribution prediction model and the object state classification prediction model which are obtained after training;
and inputting the current data stream of the state prediction object into a calculation model system to predict the state of the object, wherein the object is monitored equipment in the monitoring system.
The invention also provides an object state prediction system, which is suitable for an intelligent monitoring system for equipment operation and maintenance, and comprises the following components:
the modeling module is used for grouping monitoring data indexes of the monitored object according to the knowledge base and the associated data mining, dividing the indexes into main indexes and reference indexes, wherein the main indexes directly influence the state of the object, the reference indexes transmit the change of the reference indexes to the main indexes through the association relation, and a state analysis model of the dependency relation between the indexes and the state and the value transmission is established, and comprises the following steps: the method comprises the steps of carrying out value prediction on a reference index based on a time sequence through a single index value prediction model, a main index distribution prediction model for realizing multi-factor associated main index distribution prediction by adopting a Bayesian network, and an object state classification prediction model;
the clustering module is used for carrying out data acquisition according to the observation indexes, and clustering the monitoring sampling data or the predicted values of all indexes to form different distribution clusters; and carrying out serialization processing on the distribution clusters to obtain the distribution cluster information of the discretized sampling data or the prediction values;
the training module is used for selecting a smooth processing function according to the sample data distribution characteristics corresponding to the reference indexes, converting original data into a data sequence and bringing the data sequence into the constructed single index value prediction model for training iteration, and acquiring a stable single index value prediction model based on a time sequence; taking the distribution cluster information of the sample data of the reference index as a training value to be brought into a constructed Bayesian network representing the incidence relation between the main index and the reference index for training, and obtaining a stable multi-factor correlated main index distribution prediction model; taking the distribution cluster information of the main index prediction value as a training value to be brought into a constructed classification model for predicting and classifying object states for training to obtain a stable object state classification prediction model; a calculation model system is formed by the stable single index value prediction model, the main index distribution prediction model and the object state classification prediction model which are obtained after training;
and the prediction module is used for inputting the current data stream of the state prediction object into a calculation model system to predict the state of the object, wherein the object is the monitored equipment in the monitoring system.
The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program operable on the processor, and the processor executes the steps of the object state prediction method when executing the computer program.
The present invention also provides a storage medium storing a computer program that can be executed to perform the steps of the object state prediction method described above.
According to the embodiment of the invention, the reference index is predicted by adopting the single index value prediction model based on the time sequence in the state analysis model, and then the main index which is related to the reference index and influences the state of the object is predicted, so that the dynamic prediction of the state influence factors is realized, a practical calculation model system is formed by constructing and training the state analysis model, and a solution is provided for intelligent monitoring of mass data, so that the accuracy of object state prediction is improved, the intelligent monitoring is facilitated, and the method has positive application values in the aspects of large-scale complex equipment monitoring, traffic, logistics, smart cities, environmental protection and the like.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a flowchart of an object state prediction method according to an embodiment of the present invention;
FIG. 2 is a diagram of a technical architecture provided by an embodiment of the present invention;
FIG. 3 is a flowchart of pattern mining of index distribution according to an embodiment of the present invention;
FIG. 4 is a flow chart of state analysis model training in an embodiment of the present invention;
FIG. 5 is a flowchart illustrating a training process of a single index prediction model according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of a network structure of a distribution prediction model of a main index according to an embodiment of the present invention;
FIG. 7 is a flowchart illustrating the training of a distribution prediction model of a main index according to an embodiment of the present invention;
FIG. 8 is a diagram illustrating a network structure of a state classification prediction model according to an embodiment of the present invention;
FIG. 9 is a flowchart illustrating the training of the distribution prediction model of the main indicators according to an embodiment of the present invention;
FIG. 10 is a flowchart illustrating the process of predicting the object state according to an embodiment of the present invention;
FIG. 11 is a flowchart illustrating the calculation of object state prediction according to an embodiment of the present invention;
fig. 12 is a schematic structural diagram of an object state prediction system according to an embodiment of the present invention;
fig. 13 is a schematic structural diagram of another object state prediction system according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to make the technical solution of the present invention clearer, embodiments of the present invention are described in detail below with reference to the accompanying drawings.
The embodiment of the invention takes the operation and maintenance of massive base stations in the mobile communication industry as an application scene, and provides a calculation framework for state analysis and prediction aiming at the development requirements of intellectualization and refinement of operation and maintenance management work. In this architecture, object states are first classified according to business domain knowledge. On the basis, a three-level analysis model of 'state-main index-reference index' is formed by combining expert knowledge and association mining. Around the model, a time series analysis model is established for the reference index, and the value prediction is realized. On the basis, the value distribution prediction of the main index is realized in a mode of joint distribution probability (Bayesian network). And finally, realizing classification prediction of the object state by using a BP neural network on the basis of distribution prediction of a plurality of main index values.
Fig. 1 is a flowchart of an object state prediction method according to an embodiment of the present invention, and fig. 2 is a technical structure diagram according to an embodiment of the present invention, where the flowchart shown in fig. 1 according to the embodiment of the present invention may be implemented by the technical structure diagram shown in fig. 2. The technical structure is shown in fig. 2, and the whole computing architecture consists of three links of modeling, training mining and computing analysis. Wherein:
modeling: the main objective of this link is to establish object state classification, establish index grouping, establish numerical value transmission relationship among multiple index grouping, and finally form a connection relationship with the state classification. Thereby forming a multi-level index-state network structure. In the modeling process, firstly, on the basis of a data index set collected by a monitored object, a batch of indexes with definite characterization modes are screened out by combining means such as expert field knowledge, mechanism models and the like. On the basis, index groups are established according to related knowledge or an association mining means. In these groups, some groups have a direct value transfer relationship with respect to the object state, and these groups are called main index groups, and the groups in which other indexes are located become reference index groups. In daily operation of the reference indexes, the conversion of the values of the reference indexes does not directly influence the object state, but influences the main indexes with the association relation in the main index grouping through the transfer relation. In the invention, the prediction is realized through three-level relations of value prediction and change transmission of the reference index, main index distribution judgment and object state classification.
In the three-level model, a primary index may be associated with one or more reference indices. In general, the value distribution of the main index is obtained by calculating the predicted value of the reference index, and the main index itself can also be independently used as a single index value prediction process based on a time series to a certain extent.
Training and excavating: in this link, training of each model is achieved by bringing in sample data. Firstly training according to sample time series data of a reference index to obtain a value prediction model corresponding to the index; and then, clustering all data of the reference indexes to obtain corresponding value distribution cluster results. Thus, the predicted value can be mapped in a certain cluster, and discretization processing can be carried out on the predicted value; then, establishing a Bayesian network between the reference indexes and the corresponding main indexes, and training to obtain a main index joint probability distribution prediction model; and finally, on the basis of the plurality of main indexes, constructing a neural network, forming a state classification model based on the plurality of main indexes through sample training, and obtaining a complete calculation model system after training and mining are completed.
Calculating and analyzing: in this link, namely the state prediction link, the real-time data stream is accessed, and the state of the object at the time ti +1 is predicted at the time ti. In the process, the data corresponding to the reference index in the access data stream is processed and the value of the reference index at the time ti +1 is predicted. After the value is obtained, clustering is carried out on the value in the corresponding value distribution cluster, and a prediction distribution result is obtained. And then, substituting the prediction distribution results of all the reference indexes as input into a joint probability distribution prediction model, and calculating the maximum possible distribution range of all the main indexes at the time ti + 1. And then substituting the result as an input into the state classification model to obtain a final state prediction result, thereby realizing the state prediction of the object.
As shown in fig. 1, the object state prediction method in this embodiment includes:
In the step, an index with a definite characterization mode is screened out from a monitoring data set of a monitored object according to a knowledge base; and then, grouping the indexes by adopting an association mining method, and dividing the indexes into main indexes and reference indexes, wherein the main indexes directly influence the state of the object, and the reference indexes transmit the change of the reference indexes to the main indexes through an association relation.
The single index value prediction model is used for predicting the value of the reference index based on the time sequence, the main index distribution prediction model adopts a Bayesian network to realize multi-factor associated main index distribution prediction, and the object state classification prediction model adopts a neural network to realize classification prediction of the object state. In practical application, the number of layers in the middle layer of the neural network and the number of nodes in each layer can be flexibly configured according to the scale and/or the computational performance requirement of the access analysis object.
Because one monitored object has a plurality of observation indexes, the monitoring system realizes data acquisition based on the indexes, and in the state prediction process, the state needs to be classified and a prediction conclusion is formed by depending on the calculation of the indexes. Therefore, in the present invention, it is necessary to first establish an index-state dependency relationship and value transfer model, which is basically defined as follows:
StateModel＝{stateDef，mainParams，ReferParams}
wherein:
stateDef defines a set for the object state, where all state information of the object is defined as follows:
stateDef＝{NaN，health，risk，alert，sick}
wherein
NaN indicates that the object is currently in an unknown state, which is typically caused by insufficient data quality or lost data;
the health represents that the current state of the object is good, and the condition that the performance is degraded or the running condition reaches a boundary does not occur;
the risk represents that the current object is in a risk state, certain performance degradation occurs, or a part of main indexes overflow a safety boundary;
the mock represents that the current object is in an invalid/disabled state and cannot normally run;
from the viewpoint of the law of the development of a fault, the sequence of deterioration of the above-mentioned states is expressed as
dis(health，sick)＜dis(risk，sick)
The main index set is the mainParams in the model. The main index set comprises a group of observation indexes, and the abnormality of the index values directly causes the invalidation and the incapability of the object in a certain property aspect, which is defined as follows:
mainParams＝{(A，w，vCluster，CPs，CAs)|i＝1，2，....n}
it consists of an index group, in which:
a represents the mark of the main index;
w is the weight of the influence of the main index on the final state, and the weight value needs to be obtained through model training;
vCluster is a value domain distribution cluster set of the main index, which is formed by a group of cluster information, and the distribution cluster set needs to be obtained by mining;
CPs is a combined probability distribution table of the main index and the corresponding reference index with incidence relation. It is defined as follows:
vi∈vCluster
Ck∈ReferParams，Cvp∈Ck
CAs is a transfer relationship network in which an association relationship between a current reference index and a main index is defined as follows:
CAs＝{CP|p＝1，2，...n}
Cp∈ReferParams
this transitive relationship network is mined by expert knowledge or index association, such as: the Apriori method is implemented and constructed, and the specific method is not described in detail.
ReferParams is a reference index set, which includes a set of reference indexes, and the data of the indexes do not directly reflect the change degree of the object property. However, the fluctuation or distribution of the values in the time series affects the value distribution of the main index through the transfer relationship, and in turn, leads to the transformation of the object state. It is defined as follows:
ReferParams＝{(C，CAs)i|i＝1，2，...m}
c is a reference index, which is defined as follows:
C＝{Ckey，{Cvp|p＝1，2，...n}}
wherein, the Ckey is the name of the reference index;
Cvpa cluster set is distributed for the value of the index, and the cluster is obtained by mining.
By the method, a three-level model of 'state-main index-reference index' is constructed, and around the model, firstly, a time series analysis model is established for the reference index to realize the value prediction; secondly, value distribution prediction of the main index is realized in a combined distribution probability (Bayesian network) mode; and finally, realizing classification prediction of the object state by using a BP neural network on the basis of distribution prediction of a plurality of main index values.
After the three-level model is constructed, the index value distribution mode needs to be mined to facilitate the prediction calculation. Fig. 3 is a flowchart of pattern mining of index distribution in an embodiment of the present invention, where the pattern mining mainly refers to mining an index value distribution pattern, clustering data distribution according to daily monitoring data to form a distribution cluster, and then discretizing continuous waveform data.
And clustering the data of the indexes to form a distribution cluster and marking the distribution cluster. In the analysis process, the sampled data or the predicted value and the cluster are calculated to obtain the distribution cluster information of the sampled data or the predicted value; the discretization processing of the sampling data and the predicted value is realized through the extraction of the distributed cluster information, and the discretization processing is used as the input of the prediction model, so that the model calculation efficiency is improved.
In the present invention, pattern mining involves a main index and a reference index, a value distribution pattern of an object main index is stored in the aforementioned vccluster, and a value distribution pattern of a reference index is stored in Cvp. The two types of index value distribution mode mining adopt the same algorithm, and the specific processing process is shown in fig. 3.
fig. 4 is a flowchart of state analysis model training in the embodiment of the present invention, where the state analysis model mainly includes a single index value prediction model based on a time sequence, a multi-factor associated main index distribution prediction model, and an object state classification prediction model. The training process of these models is specifically shown in fig. 4:
202, taking the distribution cluster information of the sample data of the reference index as a training value to be brought into a constructed Bayesian network representing the incidence relation between the main index and the reference index for training, and acquiring a stable multi-factor correlated main index distribution prediction model;
and 203, taking the distribution cluster information of the main index prediction value as a training value to be brought into the constructed classification model for object state prediction classification for training, and obtaining a stable object state classification prediction model.
The model and the model training process are respectively as follows:
1. single index value prediction model based on time series and model training
Fig. 5 is a flowchart of a training process of a single-index prediction model in an embodiment of the present invention, and as shown in fig. 5, the model mainly aims at a reference index, and establishes a prediction model on the basis of a time series, and an Arima algorithm is adopted as an algorithm required by the prediction model in the present invention. In the training process, firstly, a proper smoothing method is selected according to the characteristics of reference index value transformation, the method comprises D-order difference, logics and the like, original data are converted to form a data sequence, then the data sequence is brought into a model, and after multiple iterations, a stable single index value prediction model based on a time sequence is formed.
2. Multi-factor associated main index distribution prediction model and model training
Fig. 6 is a schematic network structure diagram of a main index distribution prediction model in an embodiment of the present invention, as shown in fig. 6, the model is a bayesian network constructed based on an association relationship between a main index and a reference index, and based on a characteristic of the present invention corresponding to an object scale in an application scenario, in order to reduce a calculation amount and improve efficiency, the bayesian network employs a simple one-layer network structure as shown in fig. 6, the model establishes a prediction bayesian network for each main index, and the network is composed of an input layer and an output layer. Each node in the input layer is a reference index having an association relation with the main index, and the output layer is a joint distribution probability table corresponding to the main index. The reference indexes C1, C2 and … Cn nodes of the input layer establish connection relations with the output layer, and probability tables between the input nodes and the output nodes are recorded in the connection relations.
Fig. 7 is a training flow chart of the main index distribution prediction model in the embodiment of the present invention, as shown in fig. 7, after the model is constructed, sample data is taken in for training, in the training process, first, the sample data of C1, C2, and … Cn needs to be subjected to value processing to obtain corresponding distribution cluster information, the distribution cluster information is taken as a training value to be taken in a network, and a stable bayesian network is formed through iteration, and a specific process is shown in fig. 7.
3. Object state classification prediction model and model training
After the distribution of the predicted values of the main indexes is obtained, the states of the objects can be classified according to the predicted values of the main indexes, and then prediction is achieved. As described above, in the present invention, the state of an object is classified into several levels, i.e., NaN, health, risk, and sick, and thus, the discrete expression of state information is realized.
Fig. 8 is a schematic diagram of a network structure of a state classification prediction model in an embodiment of the present invention, and as shown in fig. 8, in the present invention, joint classification prediction of a plurality of main indexes is realized through a BP neural network with respect to characteristics of an application scenario. By combining the characteristic of large object scale in the application scene corresponding to the invention, the neural network adopts a scalable hierarchical relationship. And establishing a BP neural network aiming at object state prediction, wherein the network takes the main index prediction value distribution cluster as a network input node, each node in an input layer corresponds to a certain main index, an intermediate layer is a telescopic layer, and the number of layers and the number of nodes in each layer are elastically and telescopically configured according to the overall calculation performance requirement. In the application scenario mentioned in the invention, in order to improve the calculation efficiency, a layer structure is adopted, the number of hidden units in the layer is consistent with the number of nodes of an input layer, and an output layer is a corresponding state classification result.
Fig. 9 is a training flow chart of the main index distribution prediction model in the embodiment of the present invention, as shown in fig. 9, after the model is constructed, sample data is taken in for training, and in the training process, first, the sample data of a1, a2, and … Am needs to be subjected to value processing to obtain corresponding value distribution cluster information, and the value distribution cluster information is taken as a training value to be taken in a network, and a stable classification network is formed through iteration.
In practical application, different classification algorithms can be adopted for adaptation according to the characteristics of objects in different application scenes and the requirements of computing performance, so that flexible adaptation is realized.
And step 30, inputting the current data stream of the state prediction object into a calculation model system to predict the state of the object.
Fig. 10 is a schematic flowchart of predicting an object state in the embodiment of the present invention, and the actual state prediction is as shown in fig. 10, and includes the following steps:
301, sampling and clustering the current data stream of the monitored object, and acquiring the distribution cluster information of the current main index of the monitored object;
and step 304, bringing the distribution cluster information of the main index into the trained object state classification prediction model to obtain the prediction state of the object.
In the monitoring activity aiming at the mass objects, the intelligent analysis is realized on the states of different objects by utilizing the model and the algorithm. In the process, the application system firstly constructs a sample set by using historical data, and completes the mining of the correlation value distribution pattern and the training of a prediction model aiming at each object on the basis of the sample set to form a complete analysis knowledge base. On the basis, the incremental data are analyzed, and the recognition and early warning of the object state are realized.
Fig. 11 is a specific flowchart of the object state prediction calculation in the embodiment of the present invention, as shown in fig. 11, in the early warning process, first, the distribution cluster detection is performed on the main index value, and if the main index distribution cluster detection result of the object has no abnormal condition and has no abnormal condition after the state sequence detection, the current object state is marked as the normal state health; if the main index distribution cluster detection result of the current object is partially abnormal, extracting reference index data of the current object, bringing the reference index data into the prediction model to perform state prediction, marking state information of the current object according to the result, if the state of the object has the possibility of abnormal occurrence, marking the state of the object as a prediction result risk, and if the state of the object does not have the possibility of abnormal occurrence, marking a normal state health; and if the detection result of the main index distribution cluster of the current object has a large range abnormality, marking the current state sick and predicting the subsequent state.
In the state prediction marking process, in order to improve the working efficiency, whether the state of the object reaches an abnormal standard or not can be judged before the state is predicted, namely when the current state of the monitored object is detected to be an abnormal state and the state of the object reaches a preset abnormal standard, the state of the object can be directly marked, and prediction calculation is not needed to judge so as to save the calculated amount and the state prediction time of the system; and when the current state of the monitored object is detected to be an abnormal state and the state of the object does not reach a preset abnormal standard, predicting the state according to a prediction method adopted by the prediction model, wherein the preset abnormal standard can be that a specific part of main indexes are abnormal or that a plurality of main indexes are abnormal at the same time.
In the embodiment of the invention, in the real-time state detection process of massive objects, pattern matching is firstly carried out on a main index, the state of a current object is identified, and the subsequent analysis and calculation process is only carried out when an abnormal object is identified; and after the object state is identified to be abnormal, marking the current state of the object according to the mode matching result of the main index, acquiring the predicted state of the object according to the calculation result of the three-level model, and marking and issuing the result.
According to the embodiment of the invention, the reference index is predicted by adopting the single index value prediction model based on the time sequence in the state analysis model, and then the main index which is related to the reference index and influences the state of the object is predicted, so that the dynamic prediction of the state influence factors is realized, a practical calculation model system is formed by constructing and training the state analysis model, and a solution is provided for intelligent monitoring of mass data, so that the accuracy of object state prediction is improved, the intelligent monitoring is facilitated, and the method has positive application values in the aspects of large-scale complex equipment monitoring, traffic, logistics, smart cities, environmental protection and the like.
For the false alarm problem in the operation and maintenance system, the invention provides an object state prediction technology based on a multi-index time sequence on the basis of a big data and artificial intelligence technology. The technology screens a group of indexes on the basis of a monitoring data set, the group of indexes consists of main indexes and relevant reference indexes, the health state of a cell is obtained by joint calculation of the main indexes, the value distribution of the main indexes is influenced by the numerical value change of the reference indexes, and in the calculation process, a time series prediction function is established aiming at the reference indexes to predict the distribution interval corresponding to the value of the next time node. On the basis, the prediction value distribution interval is used as the input of a main index prediction model, so that the value prediction of a main index is realized, and the prediction of the equipment state is finally realized, thereby effectively reducing the actual condition of false alarm.
The method provided by the invention can be used for realizing health state evaluation and prediction aiming at specific objects, and is also suitable for intelligent monitoring and operation and maintenance management scenes of similar equipment such as power grids, rail transit, aircrafts and the like.
Fig. 12 is a schematic structural diagram of an object state prediction system according to an embodiment of the present invention, and as shown in fig. 12, the system of the present invention includes: the system comprises a modeling module 100, a training module 200 and a prediction module 300, wherein the modeling module 100 is used for grouping monitoring data indexes of a monitored object according to a knowledge base and associated data mining and constructing a state analysis model, and the state analysis model comprises: the system comprises a single index value prediction model, a main index distribution prediction model and an object state classification prediction model; the training module 200 is used for performing model training on the state analysis model by using the sample data to obtain a calculation model system; and the prediction module 300 is configured to input the current data stream of the state prediction object into the calculation model system to predict the state of the object.
The second embodiment of the present invention can implement the solution of the first embodiment of the method, and the working principle and the achieved technical effect are similar, and are not described again.
On the basis of the above-mentioned fig. 12, the system of the present invention further includes a clustering module for discretizing the data cluster, so as to facilitate the computer processing.
Fig. 13 is a schematic structural diagram of another object state prediction system according to an embodiment of the present invention, and as shown in fig. 13, the system according to the embodiment further includes, on the basis of the second embodiment: the clustering module 400 is used for clustering the monitoring sampling data or the predicted values of each index to form different distribution clusters; and carrying out serialization processing on the distribution clusters to obtain the distribution cluster information of the discretized sampling data or the prediction values. And (4) performing analysis prediction calculation on each model input by clustering serialization of sampling values or predicted values.
In a specific embodiment of the object state prediction system of the present invention, the modeling module 100 may be specifically configured to screen out an index with a definite characterization pattern from a monitoring data set of a monitored object according to a knowledge base; and index grouping is carried out by adopting a correlation mining method, and the indexes are divided into main indexes and reference indexes. The training module 200 is specifically configured to select a smoothing function according to sample data distribution characteristics corresponding to the reference index, convert the original data into a data sequence, bring the data sequence into the constructed single-index-value prediction model, perform training iteration, and obtain a stable single-index-value prediction model based on a time sequence; taking the distribution cluster information of the sample data of the reference index as a training value to be brought into a constructed Bayesian network representing the incidence relation between the main index and the reference index for training, and obtaining a stable multi-factor correlated main index distribution prediction model; and taking the distribution cluster information of the main index prediction value as a training value to be brought into the constructed classification model for object state prediction classification for training, and obtaining the stable object state classification prediction model. The prediction module 300 is specifically configured to sample and cluster a current data stream of a monitored object, and obtain distribution cluster information of a current main index of the monitored object; substituting the distribution cluster information of the current main index into a calculation model system, detecting the current state of the monitored object, if the monitored object is in an abnormal state, acquiring the sampling data of the reference index corresponding to the current main index of the monitored object, performing time sequence prediction on the sampling data, and acquiring the distribution cluster information of the corresponding reference index according to the predicted value; respectively bringing the distribution cluster information of the reference indexes into corresponding trained main index distribution prediction models to obtain the distribution cluster information of the main indexes; and substituting the distribution cluster information of the main index into the trained object state classification prediction model to obtain the prediction state of the object.
An embodiment of the present invention further provides a computer device, which includes a memory and a processor, where the memory stores a computer program that can be executed on the processor, and the processor executes the steps of the object state prediction method shown in fig. 1 when executing the computer program.
An embodiment of the present invention further provides a storage medium, where the readable storage medium includes: ROM/RAM, magnetic disks, optical disks, etc., and the storage medium stores a computer program that can be executed by a hardware device such as a terminal device, a computer, or a server to perform the steps of the object state prediction method described above.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (9)
1. An object state prediction method is suitable for an intelligent monitoring system for equipment operation and maintenance, and is characterized by comprising the following steps:
grouping monitoring data indexes of a monitoring object according to a knowledge base and associated data mining, dividing the indexes into main indexes and reference indexes, wherein the main indexes directly influence the state of the object, the reference indexes transmit the change of the reference indexes to the main indexes through an association relation, and a state analysis model of index-state dependency relation and value transmission is constructed, and the state analysis model comprises: the method comprises the steps of carrying out value prediction on a reference index based on a time sequence through a single index value prediction model, a main index distribution prediction model for realizing multi-factor associated main index distribution prediction by adopting a Bayesian network, and an object state classification prediction model;
carrying out data acquisition according to the observation indexes, clustering the monitoring sampling data or predicted values of all indexes to form different distribution clusters, and carrying out serialization processing on the distribution clusters to obtain the distribution cluster information of the discretized sampling data or predicted values;
selecting a smooth processing function according to sample data distribution characteristics corresponding to the reference index, converting original data into a data sequence, bringing the data sequence into the constructed single-index-value prediction model, and performing training iteration to obtain a stable single-index-value prediction model based on a time sequence;
taking the distribution cluster information of the sample data of the reference index as a training value to be brought into a constructed Bayesian network representing the incidence relation between the main index and the reference index for training, and obtaining a stable multi-factor correlated main index distribution prediction model;
taking the distribution cluster information of the main index prediction value as a training value to be brought into a constructed classification model for predicting and classifying object states for training to obtain a stable object state classification prediction model;
a calculation model system is formed by the stable single index value prediction model, the main index distribution prediction model and the object state classification prediction model which are obtained after training;
and inputting the current data stream of the state prediction object into a calculation model system to predict the state of the object, wherein the object is monitored equipment in the monitoring system.
2. The method according to claim 1, characterized in that the monitoring data indicators of the monitored objects are grouped according to a knowledge base and associated data mining, specifically:
screening indexes with definite characterization modes from a monitoring data set of a monitored object according to a knowledge base;
and (5) performing index grouping by adopting a correlation mining method.
3. The method of claim 1, wherein the object state classification prediction model uses a neural network to perform classification prediction of the object state.
4. The method of claim 3, wherein the number of layers in the middle layer of the neural network and the number of nodes in each layer are flexibly configurable according to the scale and/or computational performance requirements of the access analysis object.
5. The method according to any one of claims 1 to 4, wherein the current data stream of the state prediction object is input into a computational model system, and the state of the object is predicted, specifically:
sampling and clustering the current data stream of the monitored object to obtain the distribution cluster information of the current main index of the monitored object;
substituting the distribution cluster information of the current main index into a calculation model system, detecting the current state of the monitored object, if the monitored object is in an abnormal state, acquiring the sampling data of the reference index corresponding to the current main index of the monitored object, performing time sequence prediction on the sampling data, and acquiring the distribution cluster information of the corresponding reference index according to the predicted value;
respectively bringing the distribution cluster information of the reference indexes into corresponding trained main index distribution prediction models to obtain the distribution cluster information of the main indexes;
and substituting the distribution cluster information of the main index into the trained object state classification prediction model to obtain the prediction state of the object.
6. The method according to claim 5, wherein when the current state of the monitored object is detected to be abnormal and the object state reaches a preset abnormal standard, the object state is directly marked.
7. An object state prediction system is applicable to an intelligent monitoring system for equipment operation and maintenance, and is characterized by comprising:
the modeling module is used for grouping monitoring data indexes of the monitored object according to the knowledge base and the associated data mining, dividing the indexes into main indexes and reference indexes, wherein the main indexes directly influence the state of the object, the reference indexes transmit the change of the reference indexes to the main indexes through the association relation, and a state analysis model of the dependency relation between the indexes and the state and the value transmission is established, and comprises the following steps: the method comprises the steps of carrying out value prediction on a reference index based on a time sequence through a single index value prediction model, a main index distribution prediction model for realizing multi-factor associated main index distribution prediction by adopting a Bayesian network, and an object state classification prediction model;
the clustering module is used for collecting data according to the observation indexes, clustering the monitoring sampling data or the predicted values of all indexes to form different distribution clusters; and carrying out serialization processing on the distribution clusters to obtain the distribution cluster information of the discretized sampling data or the prediction values;
the training module is used for selecting a smooth processing function according to the sample data distribution characteristics corresponding to the reference indexes, converting original data into a data sequence and bringing the data sequence into the constructed single index value prediction model for training iteration, and acquiring a stable single index value prediction model based on a time sequence; taking the distribution cluster information of the sample data of the reference index as a training value to be brought into a constructed Bayesian network representing the incidence relation between the main index and the reference index for training, and obtaining a stable multi-factor correlated main index distribution prediction model; taking the distribution cluster information of the main index prediction value as a training value to be brought into a constructed classification model for predicting and classifying object states for training to obtain a stable object state classification prediction model; a calculation model system is formed by the stable single index value prediction model, the main index distribution prediction model and the object state classification prediction model which are obtained after training;
and the prediction module is used for inputting the current data stream of the state prediction object into a calculation model system to predict the state of the object, wherein the object is the monitored equipment in the monitoring system.
8. A computer device comprising a memory and a processor, the memory having stored thereon a computer program operable on the processor, wherein the processor, when executing the computer program, performs the steps of the object state prediction method of any one of claims 1 to 6.
9. A storage medium, characterized in that the storage medium stores a computer program that can be executed to perform the steps of the object state prediction method according to any one of claims 1 to 6.
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 CN109872003A (en) | 2019-06-11 |
| CN109872003B true 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) |
Families Citing this family (11)
| 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 |
| CN110531988B (en) * | 2019-08-06 | 2023-06-06 | 新华三大数据技术有限公司 | Application program state prediction method and related device |
| 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 |
| 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 |
| CN112052914B (en) * | 2020-09-29 | 2023-12-01 | 中国银行股份有限公司 | Classification model 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 |
| 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 |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN105760970A (en) * | 2016-03-21 | 2016-07-13 | 重庆灵狐科技股份有限公司 | Method for predicting AQI |
| 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 |
Family Cites Families (9)
| 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 |
| 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 |
| CN105468917B (en) * | 2015-12-01 | 2019-03-08 | 北京无线电计量测试研究所 | A kind of pipeline failure prediction technique and device |
| 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 |
| CN107991870B (en) * | 2017-12-05 | 2020-10-27 | 暨南大学 | Fault early warning and service life prediction method for escalator equipment |
| CN108020781B (en) * | 2017-12-19 | 2020-05-29 | 上海电机学院 | Breaker fault diagnosis 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 |
| 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 (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN105760970A (en) * | 2016-03-21 | 2016-07-13 | 重庆灵狐科技股份有限公司 | Method for predicting AQI |
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109872003A (en) | 2019-06-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109872003B (en) | Object state prediction method, object state prediction system, computer device, and storage medium | |
| CN105512448A (en) | Power distribution network health index assessment method | |
| WO2019196869A1 (en) | Method for determining list of patrolling base stations, and patrolling apparatus | |
| CN105325023A (en) | Method and network device for cell anomaly detection | |
| CN110977614A (en) | Health diagnosis method for numerical control machine tool | |
| CN115423009A (en) | Cloud edge coordination-oriented power equipment fault identification method and system | |
| CN116865258A (en) | Hierarchical distributed power supply intelligent power grid construction method | |
| JP2023504103A (en) | MODEL UPDATE SYSTEM, MODEL UPDATE METHOD AND RELATED DEVICE | |
| CN103617447A (en) | Evaluation system and method for intelligent substation | |
| CN116307641A (en) | Digital power plant-oriented resource collaborative scheduling management method and system | |
| CN116308304A (en) | New energy intelligent operation and maintenance method and system based on meta learning concept drift detection | |
| CN117494009A (en) | Electrical equipment state evaluation method based on insulating material pyrolysis analysis and cloud platform | |
| CN114936808A (en) | Cloud-edge cooperative task management system and method for substation fault detection | |
| CN115905348A (en) | Industrial electricity utilization abnormity early warning method and system of multi-source heterogeneous data | |
| CN113902164A (en) | Power distribution network fault prediction system and method based on big data technology | |
| Wang et al. | LSTM-based alarm prediction in the mobile communication network | |
| Jingyu et al. | Statistical analysis of distribution network fault information based on multi-source heterogeneous data mining | |
| Zheng et al. | [Retracted] Application Based on Artificial Intelligence in Substation Operation and Maintenance Management | |
| Tan et al. | Research on Fault Prediction Model Based on 5G Data Center | |
| Mavi et al. | On time demand traffic estimation based on DBN with Horse herd optimization for next generation wireless network | |
| Xiaodong et al. | Artificial Intelligence in power multimodal data analysis | |
| LIU et al. | A case study on intelligent operation system for wireless networks | |
| Hu et al. | A Bad Data Identification Method for Multiple Spatio-temporal Data in Power Distribution Network | |
| US20230315955A1 (en) | System and method for optimizing utility pipe sensors placement using artificial intelligence technology | |
| CN117994076A (en) | Natural gas pipe network scheduling energy-saving assessment method based on big data |
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 |