CN116596405B - Pure data driven power system analysis method and system - Google Patents

Abstract

The application relates to a pure data-driven power system analysis method and system. The method comprises the following steps: responding to a service processing request of a service end, and determining data query information corresponding to the service processing request; obtaining target power data according to data query information and a pre-constructed knowledge graph index library, wherein the knowledge graph index library is an index library generated based on a pre-constructed unified information model, and the unified information model is a model generated by expanding based on a public information model; and mapping the target power data to a virtual form, wherein the virtual form is a form which can be directly used when the service end performs service processing, and is used for indicating the service end to perform service processing. By adopting the method, the data barriers can be opened, and the multi-source data of the novel power system are fused, so that the novel power system can rapidly process the service processing request initiated by the service end in a pure data driving mode.

Description

Pure data driven power system analysis method and system

Technical Field

The present application relates to the field of data processing technology, and in particular, to a pure data driven power system analysis method, apparatus, computer device, storage medium, and computer program product.

Background

The novel power system has the characteristics of high construction speed and more system nodes. At present, under the operation scene of a novel power system, because the structure of a power network is staggered and complicated, the types of power users are various, and the storage mode, the data standard and the like adopted by data accumulated in the development process of the power system are not uniform, so that the data of the power system show a multi-source development trend.

In the conventional art, it is generally necessary to write a script in advance, and when a task of a data processing service side needs to be invoked, the script is executed to pull data from different databases, respectively. With the rapid development of the novel power system, the service processing is performed by using the mode in the traditional technology, which easily results in insufficient service processing efficiency.

Disclosure of Invention

In view of the foregoing, it is desirable to provide a pure data driven power system analysis method, system, apparatus, computer device, computer readable storage medium, and computer program product that can improve the efficiency of power system business processes.

In a first aspect, the present application provides a pure data driven power system analysis method, the method comprising:

Responding to a service processing request of a service end, and determining data query information corresponding to the service processing request;

obtaining target power data according to the data query information and a pre-constructed knowledge graph index library, wherein the knowledge graph index library is an index library generated based on a pre-constructed unified information model, and the unified information model is a model generated by expanding based on a public information model;

mapping the target power data to a virtual table, wherein the virtual table is a table which can be directly used when business processing is carried out;

and carrying out service processing based on the virtual form to obtain a service processing result.

In one embodiment, the generating manner of the unified information model includes:

under the condition that the power data are analyzed and the first equipment class which does not meet modeling requirements exists, similarity between the first equipment class and the second equipment class existing in the public information model is determined;

generating an expansion model according to the similarity, the first equipment class and attribute information corresponding to the first equipment class;

and generating the unified information model according to the extension model and the public information model.

In one embodiment, the generating an extension model according to the similarity, the first device class, and attribute information corresponding to the first device class includes:

when a second equipment class similar to the first equipment class exists in the public information model based on the similarity, creating a new class according to the similar second equipment class, constructing an inheritance relationship between the new class and the similar second equipment class, and generating the expansion model based on the new class and attribute information corresponding to the new class;

when it is determined that a second equipment class similar to the first equipment class does not exist in the public information model based on the similarity, a new class is created according to the first equipment class, and the expansion model is generated based on the new class and attribute information corresponding to the new class.

In one embodiment, the obtaining the target power data according to the data query information and the pre-constructed knowledge-graph index library includes:

determining a target element corresponding to the data query information;

and generating a query statement according to the target element, executing the query statement, and querying from the pre-constructed knowledge graph index library to obtain power data corresponding to the target element as the target power data.

In one embodiment, the business process request is a risk assessment request, and the virtual form includes power load data required for processing the risk assessment request; after mapping the target power data to the virtual table, the method further includes:

directly using the power load data in the virtual table to respectively generate a risk index and an element failure rate of the power system;

and taking the risk index, the element fault rate and the power load data as input data, and predicting by adopting a hybrid neural network to obtain power operation data and risk grade.

In a second aspect, the present application also provides a pure data driven power system analysis system, the system comprising:

the edge gateway is connected with the equipment object and is used for collecting original power data of the equipment object;

the edge cluster is in communication connection with the edge gateway and is used for responding to a service processing request of a service end and determining data query information corresponding to the service processing request; obtaining target power data according to the data query information and a pre-constructed knowledge graph index library; mapping the target power data to a virtual form, and performing service processing based on the virtual form to obtain a service processing result; the knowledge graph index library is an index library generated based on a pre-constructed unified information model, the unified information model is a model generated by expanding based on a public information model, and the virtual table is a table which can be directly used when the edge cluster performs service processing;

And the cloud system is in communication connection with the edge cluster and is used for receiving the service processing result sent by the edge cluster and performing operation control based on the service processing result.

In a third aspect, the present application also provides a pure data driven power system analysis device, the device comprising:

the query information analysis module is used for responding to the service processing request of the service end and determining data query information corresponding to the service processing request;

the data query module is used for obtaining target power data according to the data query information and a pre-built knowledge graph index library, wherein the knowledge graph index library is a graph index library generated based on a pre-built unified information model, and the unified information model is a model generated by expansion based on a public information model;

the data virtualization module is used for mapping the target power data to a virtual table, wherein the virtual table is a table which can be directly used when business processing is carried out;

and the service processing module is used for carrying out service processing based on the virtual form to obtain a service processing result.

In a fourth aspect, the present application further provides a computer device, the computer device including a memory and a processor, the memory storing a computer program, the processor implementing the method according to any one of the embodiments of the first aspect when executing the computer program.

In a fifth aspect, the present application further provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method according to any one of the embodiments of the first aspect.

In a sixth aspect, the present application also provides a computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of the embodiments of the first aspect.

According to the pure data-driven power system analysis method, the system, the device, the computer equipment, the storage medium and the computer program product, the pure data-driven power system analysis system is adopted, the unified information model, the knowledge graph index library and the data virtualization technology are utilized, the data barriers are opened, the multi-source data of the novel power system are fused, and therefore the method can be better adapted to the characteristics that the novel power system has more nodes, high system construction speed and unknown network parameter information of a plurality of power systems, and the novel power system can rapidly process service processing requests initiated by service ends in a pure data-driven mode.

Drawings

FIG. 1 is a block diagram of the overall architecture of a pure data driven power system analysis system in one embodiment;

FIG. 2 is a schematic diagram of cloud, edge, end data flow interactions in one embodiment;

FIG. 3 is a flow chart of a method of analyzing a pure data driven power system according to one embodiment;

FIG. 4 is a schematic diagram of an extension of a distributed photovoltaic device in one embodiment;

FIG. 5 is a class association diagram formed by expanding new energy equipment classes by a public information model in one embodiment;

FIG. 6 is a flow diagram of a method for generating a unified information model in one embodiment;

FIG. 7 is a flow diagram of a method for generating a unified information model in one embodiment;

FIG. 8 is a schematic diagram of a three-layer risk assessment index for an electrical power system according to one embodiment;

FIG. 9 is a schematic diagram of a distributed control architecture under cloud edge collaboration in one embodiment;

FIG. 10 is a block diagram of a pure data driven power system analysis device in one embodiment;

FIG. 11 is an internal block diagram of a computer device in one embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.

The analysis method of the pure data driven power system provided by the embodiment of the application can be applied to the analysis system of the pure data driven power system (hereinafter referred to as an "analysis system") shown in fig. 1, and referring to fig. 1, an information flow framework supporting the analysis system driven by the pure data presents a two-stage fusion architecture of a cloud end and an edge end, the cloud end deploys the cloud end system facing the full network service, and the edge end comprises a plurality of edge computing nodes including edge clusters, virtual edge clusters and edge gateway three types of edge end systems. As shown in fig. 1:

the cloud system is deployed on the scheduling cloud, and provides basic data storage, data processing, artificial intelligent computing and other resources by the lower-layer basic resources, and is in butt joint with the large data center platform upwards, and the large data center platform provides platform services for upper-layer applications (such as running control) and the like.

The edge cluster is deployed at the dispatching master station of each hierarchical network, province and region. The cloud computing resources of the provincial edge cluster can be used as cloud centers, and the territories, including substation edge gateways, can be used as edge side computing resources relative to the cloud centers. In the cloud edge collaborative architecture of the range, data virtualization is deployed at an edge cluster, a unified information model is provided for the whole network, and data service is provided through virtualization into upper-layer simulation calculation and distributed regulation. Cloud edge cooperative interaction and cloud edge cooperative application are deployed in the edge cluster. The cloud side cooperative interaction is used for achieving the functions of data interaction between the cloud system and the edge cluster, localization of cloud service and the like. The cloud edge cooperative application provides the safety access of the edge gateway, so that the edge cluster can perform the functions of monitoring, predicting, analyzing and controlling the regulation object based on the power data of the terminal equipment collected by the edge gateway.

The edge gateway is deployed at the terminal equipment nearby, provides an entrance for accessing the mass regulation and control objects into the overall distributed regulation and control architecture for the upper access edge cluster, and provides functions of cloud-edge cooperative interaction, model expansion, in-situ decision and the like by adopting a unified technical architecture.

In the whole analysis system, an information interaction mechanism for storing different business objects facing to a source network load is adopted, and referring to a cloud, side and end data flow interaction schematic diagram shown in fig. 1 and 2, a main information interaction data flow comprises the following parts:

(1) Data streams of longitudinal interactions between edge clusters and centralized plant sites (including traditional power plants, centralized new energy plant sites, etc.). The front-end system of the dispatching master station is generally communicated with an edge gateway (intelligent remote machine), original power data of terminal equipment at a station end are collected, and an operation strategy, a control instruction and the like are issued.

(2) And the data flow is longitudinally interacted between the edge cluster and mass new energy equipment, market main bodies and the like. The security access area is generally communicated with the edge gateway, original power data of new energy equipment, market main body and the like are collected, data or application service, configuration are provided, and operation strategies, control instructions and the like are issued.

(3) And the cloud system and the edge clusters (including the virtual edge clusters) longitudinally interact data streams. The cloud side interaction module of the cloud system is generally communicated with the cloud side cooperative interaction module of the edge cluster, so that the requirements, operation data and the like of the edge cluster are collected, and data services, application services, cooperative control strategies and the like are issued.

For interactions between the cloud system and the edge clusters, the interaction model is not limited to include: conventional grid models (e.g., traditional power generation, transmission, transformation, etc.) and new power system models (e.g., centralized new energy, distributed new energy, virtual power plants, integrated parks, charging facilities, etc.).

The interaction data is not limited to including: the edge cluster is used for uploading real-time data such as telemetry, remote signaling, remote pulse and the like of conventional power grid equipment, novel power system equipment and the like or data such as E-format telemetry, remote signaling and the like; the cloud system transmits control information such as remote control, remote adjustment and the like which face conventional power grid equipment, novel power system equipment and the like to the edge cluster; files such as the state of a sign board of a main distribution network, new energy equipment and the like; device status and device alarm information of conventional power grid devices, novel power system devices and the like.

For edge clusters to interact with edge gateways, the interaction model is not limited to include: the power grid model interaction comprises modeling of primary power grid equipment, novel new energy power equipment, charging piles and the like, and modeling is carried out according to hierarchical relations of substation models (plant stations, voltage levels, intervals, equipment, topology and the like); the equipment model interaction refers to modeling of an edge gateway and secondary equipment accessed by the edge gateway, and comprises association relation, measurement information, running state and the like of the secondary equipment, and basic support of application such as equipment situation awareness, online monitoring and the like is realized.

The interaction data is not limited to including: edge gateway local settings, local logs, and historical data. The method comprises the steps of supporting to receive and process power grid running state data sent by an edge gateway; supporting to receive the power data of the conventional power grid equipment and the novel power system equipment which are uploaded by the edge gateway; the support command issuing comprises instructions of switching the operation mode of the power generation network, frequency modulation, peak shaving, planning value and the like under the edge cluster.

For the edge gateway to interact with the terminal device, the interaction data is not limited to include: the power data (including data such as power grid running state and device running condition) and control instructions (including switching, frequency modulation and peak regulation of the power grid running mode under the intelligent equipment) obtained from the terminal equipment.

In a specific implementation, when a service end initiates a service processing request, an edge cluster responds to the service processing request to determine data query information corresponding to the service processing request; obtaining target power data according to data query information and a pre-built knowledge graph index library, wherein the knowledge graph index library is an index library generated based on a pre-built unified information model and original power data of each terminal device collected by an edge gateway, and the unified information model is a model generated by expansion based on a common information model CIM; the edge cluster maps the target power data to a virtual form, wherein the virtual form is a form which can be directly used when the service processing is performed, so that the service processing is performed based on the virtual form, a service processing result is obtained, and the obtained service processing result is sent to the cloud system, so that the cloud system performs operation control based on the service processing result.

The terminal device may be, but not limited to, various distributed new energy devices, transformer stations, centralized plant stations, virtual power plants, distribution stations, etc. The edge cluster and the cloud system can be implemented by independent servers or a server cluster formed by a plurality of servers.

In one embodiment, as shown in fig. 3, a pure data-driven power system analysis method is provided, and the method is applied to the edge clusters in fig. 1 and fig. 2 for illustration, and includes the following steps S310 to S340:

step S310, responding to the service processing request of the service end, and determining the data query information corresponding to the service processing request.

Wherein the traffic handling request may be, but is not limited to, a forecast request for power flow, a risk assessment request, a distributed regulation request, etc. The data query information may include, but is not limited to, device identification of the terminal device, attribute information (e.g., capacity, current), etc. The data query information can be carried in the service processing request, and can also be obtained by analyzing and processing the service processing request.

Step S320, obtaining target power data according to the data query information and a pre-constructed knowledge-graph index library.

The knowledge graph index library is an index library generated based on a pre-constructed unified information model. One implementation of the knowledge-graph index library is described below:

firstly, a unified information model can be generated by expanding based on the public information model, for example, the public information model is expanded to adapt to new energy equipment in the face of future large-scale access to the distributed new energy equipment, so as to obtain the unified information model. Fig. 4 illustrates a schematic diagram taking an extension of a distributed photovoltaic device as an example, and referring to fig. 4, a parent class of a new energy device may be designed to include sub-classes of photovoltaic modules, inverters, energy storage devices, and the like, and class attributes include device basic information, electrical parameters, operation data, and the like, and the class attributes and relationship design thereof conform to the CIM standard.

And then, combing out a knowledge graph ontology model required for constructing the knowledge graph according to the unified information model modeling document, wherein the knowledge graph ontology model comprises information such as the category of the defined entity object, attribute information contained in the entity object, the relationship among the entities and the like. Fig. 5 schematically shows a unified class association diagram formed by expanding new energy device classes by a common information model. Referring to fig. 5, the knowledge graph ontology model determines, based on this, that the entity object class is the object class of each device (terminal device) of the power grid, and the relationship between them also follows the association logic in fig. 5. And constructing key elements of the knowledge graph by combining the attribute description specifications of all the equipment types in the rule document: entity, relationship, attribute. Of course, apart from the new energy equipment class, the entity object of other electric power systems also refers to the above to build the ontology model.

Finally, the unified information model is used for unified modeling of original business data of each specialty of the company from the enterprise level perspective, and mainly comprises a logic model and a physical model, wherein the design range does not comprise various statistical data, index data, system management data and the like. The logic model is used for uniformly defining the core business objects of the company, the attribute fields and the interrelationships thereof from the enterprise-level view point, and mainly avoiding repeated definition of the same business object in the construction process of a business information system so as to further cause repeated input and maintenance. The logic model is irrelevant to organization setting, department responsibility division and specific management status quo, and is irrelevant to database products and business applications. The physical model follows a logical model, and forms a database structure and a table design based on the selected database product according to data processing requirements of the business application (such as business transaction, off-line analysis, real-time calculation and the like). The physical model is a specific floor application of the logic model, is used for supporting the requirements of business applications on storage, transmission, access, calculation and the like of data, and mainly comprises a data center sharing layer physical model and a business application information system physical model.

The physical model database, i.e. the database used for model data storage, can be realized by adopting Neo4j (a high-performance, non-relational graphic database) technology, and the physical model is transformed according to the actual mapping by taking the entity, the attribute and the relation among the entities as the basis, so that the object-oriented mode is represented. After the knowledge graph is generated based on the unified information model, a hash index, a full text index, a label index and other modes can be adopted to respectively establish a mapping relation between the entity and each business database, a mapping relation between attribute information and each business database, and a mapping relation between the entity and each business database, so as to obtain the knowledge graph index base.

Specifically, after the edge cluster acquires the data query information, a target element corresponding to the data query information is determined, wherein the target element can be any one or more of an entity, attribute information and a relationship among the entities. Metadata information corresponding to the target element is acquired, wherein the metadata information comprises a data storage position, a format and the like. Generating a query statement based on the target element and the metadata information, and querying a pre-constructed knowledge-graph index library by using the generated query statement to obtain power data corresponding to the target element as target power data.

In one example, the data query information is a device object. The edge cluster can search a target entity object corresponding to the device object and acquire metadata information corresponding to the target entity object. And generating a Cypher query statement based on the target entity object and the metadata information, executing the Cypher query statement, and searching from a knowledge graph index library to obtain target power data.

Further, in one embodiment, after the Cypher query statement is generated, optimization means such as multi-query statement merging, query filter rearrangement, query rewrite and the like can be adopted to accelerate the query efficiency of the knowledge graph index library.

In step S330, the target power data is mapped to a virtual table, which is a table that can be directly used when performing the business process.

Data virtualization, among other things, is a technique that can provide a unified, abstract, and packaged perspective to data consumers to query and process data stored in a heterogeneous data store collection. Data virtualization provides a unified view of data, and data consumers are blind and unaware that the data they acquire is from multiple data stores. Data virtualization hides the fact that data is integrated together to form a unified view.

Specifically, a data virtualization tool is arranged in the edge cluster in advance, after target power data is obtained through inquiry, the data virtualization tool is called, the target power data is mapped to a virtual table, and the data format, the data granularity and the like in the virtual table can be directly suitable for service processing.

And step S340, performing service processing based on the virtual form to obtain a service processing result.

Specifically, the edge cluster extracts the features required by processing the service processing request from the virtual table, and adopts the algorithm strategy which is deployed in advance and corresponds to the service processing request to process the service, so as to obtain the service processing result.

Further, the edge cluster may send the service processing result to the connected cloud system, so that the cloud system performs operation control based on the service processing result.

According to the analysis method of the pure data driven power system, the pure data driven analysis system is adopted, the unified information model, the knowledge graph index library and the data virtualization technology are utilized, the data barriers are opened, and the multi-source data of the novel power system are fused, so that the analysis method can be better adapted to the characteristics of multiple nodes of the novel power system, high system construction speed and unknown network parameter information of a plurality of power systems, and the novel power system can rapidly process service processing requests initiated by service ends in a pure data driven mode.

In one embodiment, as shown in fig. 6, the generation manner of the unified information model may be implemented by the following steps:

in step S610, in the case where the power data is analyzed to determine that there is a first device class that does not satisfy the modeling requirement, a similarity between the first device class and a second device class that already exists in the common information model is determined.

Specifically, if the power data is analyzed by adopting the common information model based on IEC61970, and it is determined that the first equipment class which does not meet the modeling requirement exists, the similarity between the first equipment class and each second equipment class in the existing common information model can be calculated. The similarity may be obtained by direct comparison, distance similarity (e.g., euclidean distance, cosine similarity), and the like. If the power data is analyzed by adopting the public information model based on IEC61970, and the first equipment class meeting the modeling requirement is determined to exist, the first equipment class can be directly used for generating the expansion model.

Step S620, generating an expansion model according to the similarity, the first equipment class and the attribute information corresponding to the first equipment class.

Step S630, a unified information model is generated according to the extension model and the public information model.

Specifically, referring to fig. 7, a threshold may be set in the edge cluster, the similarity corresponding to each second device class in the public model is compared with the threshold, if no similarity greater than or equal to the threshold exists, it is determined that no second device class similar to the first device class exists in the public information model, and step S7202 is continuously performed; if the similarity is greater than or equal to the threshold, it is determined that a second device class similar to the first device class exists in the public information model, and step S7204 is continuously performed.

Step S7202, creating a new class according to the first device class.

Specifically, a new name may be created based on the first device class, a new class is obtained, and step S7208 is continued.

Step S7204, creating a new class according to the similar second device class.

Specifically, in the case where the number of similar second device classes is one, naming may be performed based on the one second device class, resulting in a new class. In the case that the number of similar second device classes is plural, naming may be performed based on the second device class with the highest similarity, and a new class may be obtained. In one example, the name of the second class of devices may be directly employed for naming. In another example, a new name may be derived based on the name of the second device class, resulting in a new class.

In step S7206, after creating the new class based on the second device class, an inheritance relationship between the second device class and the new class can also be created. By adopting the method, the association relation among the classes in the expansion model can be perfected.

In step S7208, new attribute information including an internal attribute and an external attribute is added.

Specifically, in the case where a new class is created based on the first device class, naming may be performed based on attribute information of the first device class. In the case where a new class is created based on the second device class, naming may be performed based on the name of the attribute information of the second device class. In one example, the name of the attribute information of the second device class may be directly employed for naming. In another example, a new name may be derived based on the name of the attribute information of the second device class, resulting in the name of the attribute information of the new class.

Step S7210, analyzing whether the internal attribute and the external attribute need to establish an additional relationship, if so, executing step S7212; if not, step S7214 is performed.

Step S7212, analyzing the internal attribute and the external attribute of the new class, establishing the association relationship between the internal attribute and the external attribute and the new class, and defining the information such as the names, the storage formats and the like of the internal attribute and the external attribute. Step S7214 is continuously performed.

Step S7214, determining that the new class meets the modeling requirement.

Step S7216, generating an expansion model by using the new class and the attribute information of the new class.

Further, after the expansion model is generated, an association relationship between equipment class and attribute information between the expansion model and the public information model can be established, so that a unified information model is generated.

Further, after the unified information model expansion is detected, the knowledge graph index library can be updated by adopting the unified information model obtained by expansion, so that the accuracy of data in the knowledge graph index library is ensured.

In the embodiment, an expandable public information model is adopted, so that the system can be suitable for accessing new equipment (such as massive new energy equipment) in the future, and has good expandability, and the service processing capacity of the system is improved.

In one embodiment, a specific method for analyzing a pure data-driven power system is provided, which is applied to the analysis system shown in fig. 1, and includes the following steps S802 to S818:

in step S802, the edge cluster, when analyzing the power data and determining that there is a first device class that does not meet the modeling requirement, determines a similarity between the first device class and a second device class that is already present in the public information model.

In step S804, the edge cluster generates an extension model according to the similarity, the first device class, and attribute information corresponding to the first device class.

In step S806, the edge cluster generates a unified information model according to the extension model and the public information model.

Step S808, the edge cluster generates a knowledge graph based on the unified information model, and constructs a mapping relation between the entity and each service database, and a mapping relation between the attribute information and each service database, so as to generate a knowledge graph index base. The power data in each business database comprises data obtained by collecting original power data of equipment objects through an edge gateway, and performing operations such as data cleaning and data formatting on the original power data.

In step S810, the edge cluster determines, in response to the service processing request of the service end, data query information corresponding to the service processing request, where the data query information may be any one or more of an entity, attribute information, and a relationship between entities.

And step S812, obtaining target power data by the edge cluster according to the data query information and a pre-constructed knowledge-graph index library.

In step S814, the edge cluster invokes the data virtualization tool to map the target power data to a virtual table.

In step S816, the edge cluster performs service processing based on the virtual table, to obtain a service processing result.

In step S818, the edge cluster may send the service processing result to the cloud system, so that the cloud system performs operation control based on the service processing result. For example, if the risk index of a certain terminal device is greater than a set value as a result of service processing, the cloud system may generate alarm information, and send the alarm information to the terminal device through the gateway.

Several examples of business processes based on the above-described power system analysis method are shown below.

In one embodiment, the power system analysis method may be used to predict the power flow, and then the service processing request may be a predicted request of the power flow, which is specifically implemented by the following contents:

(1) And (3) data acquisition: based on the unified information model, the knowledge graph index library and the data virtualization tool described in the above embodiment, a virtualization table suitable for power flow prediction is generated. As described in the above embodiments, the data of all the different service systems have been subjected to unified information modeling and multi-source data fusion, and the virtual table in this embodiment is the data that can be directly applied to the load flow calculation algorithm.

(2) And (3) data characteristic extraction: features required for load flow calculation, such as correlations between power system nodes, time series features, periodic features, and the like, are extracted from the generated virtual table. These features may be obtained by statistical analysis, time-frequency analysis, machine learning, and the like. In this embodiment, node information of each node may be extracted from the virtual table, including active load and reactive load of the PQ node, active output and node voltage amplitude of the PV node, and node voltage amplitude and phase angle of the balance node.

(3) Load flow calculation and prediction: and carrying out power flow calculation and prediction on the extracted real-time data by using the trained neural network model, and predicting to obtain parameters such as voltage, power, current, phase angle and the like of each node in the power system.

Further, the power flow change and the problems possibly occurring in the power system can be intelligently analyzed based on parameters output by the neural network model.

Furthermore, by adopting the method of the embodiment, not only the base state power flow can be calculated, namely, the node information of each node in real time is input at a certain moment, but also the base state power flow at the moment is obtained through the prediction of the neural network model. Meanwhile, the method can also be used for dynamic (non-ground state) power flow, namely, parameter values in node information of each node are adjusted, for example, voltage, active and reactive information of certain specific power nodes are input into a neural network model, and the adjusted node information of each node is predicted to obtain power flow data. Or changing the topological structure of the system, inputting the node information of each node obtained after the topological structure is changed into a neural network model, and predicting to obtain tide data. The dynamic power flow calculation is a power flow calculation mode which is more needed in the service than the ground state power flow, and the possible distribution condition of the system power flow can be estimated better under the conditions of load fluctuation and topology change.

In one embodiment, the neural network model may employ a full connection depth neural network that includes multiple hidden layers (e.g., two layers). And in the training stage, the fully-connected deep neural network learns a large amount of historical data information in the power system, and a nonlinear model of the node load and the system power flow of the power system is obtained by fitting. After training is completed, the trained model may be validated and evaluated using a validation dataset, such as using Root Mean Square Error (RMSE), mean Absolute Error (MAE), etc., to evaluate the model's ability to check the model's accuracy and generalization ability.

In another embodiment, the above power system analysis method may be used to perform static stability analysis, in this embodiment, the service processing request is a risk assessment request, the virtual table is a table that is generated based on the unified information model, the knowledge graph index library and the data virtualization tool described in the above embodiment and is suitable for static stability analysis, and the virtual table includes power load data required for processing the risk assessment request, specifically implemented by:

in step S910, the risk index and the element failure rate of the power system are respectively generated directly using the power load data in the virtual table.

(1) Establishing a static security risk assessment index system of a power system

The safety risk assessment of the power system is mainly aimed at typical faults of the power system with high probability and low risk, and is measured by indexes such as load loss probability, power shortage expectation and the like. In fact, in recent years, extreme natural disasters occur, and although the disasters occur with strong randomness and small probability, once the disasters occur, serious power system faults are caused, and serious inconvenience is brought to daily production and life of people. The elastic evaluation of the electric power system is to study the influence of such small-probability and high-risk disasters, and mainly study the system state conversion and the capability of recovering to a stable state by a system control means when typhoons, hail, earthquakes, floods and other extreme natural disasters occur. Compared with the traditional power system risk assessment, the power system elasticity assessment not only considers the load loss of the system, but also comprehensively considers the recovery time and the recovery capacity of the system.

As shown in fig. 8, in this embodiment, aiming at the characteristics of a pure data-driven power grid static operation risk early warning model, three risk layer evaluation indexes of the power system with progressive relationship are constructed, and risk indexes, change risk indexes and comprehensive risk indexes are calculated respectively.

(a) First layer evaluation index

The power load data comprise node voltage amplitude values of all the actually measured nodes, voltage out-of-limit risk indexes and line out-of-limit risk indexes of the whole power system are obtained based on the node voltage amplitude values through calculation and are recorded as node voltage out-of-limit risk indexes (Node Voltage Violation Risk, RNVV) and line power overload risk indexes (Line Overload Risk, RLO), and the power load data can be realized through the following formulas:

wherein,representing node voltage out-of-limit risk indexes of the whole power system; />Indicate->The per unit value of the node voltage amplitude of each node; />Indicate->Severity of individual node voltage out-of-limit, when the node voltage amplitude is in the normal interval, the severity is 0, the closer to the safety limit, the greater the severity.

The line power overload risk index of the whole power system is calculated by adopting the following formula:

wherein,representing a line power overload risk index of the whole system; />Indicate->Load rate of the lines; />Indicate->Severity of line overload, when the load factor is less than 80%, the severity is 0, and the closer the load factor is to the rated load factor of the line, the greater the severity is.

(b) Second layer evaluation index

The first-layer risk assessment index is easily affected by the load level, and when the power system is at a higher load level, the calculation result of the first-layer risk index is relatively high, so that the risk condition of the power system cannot be well reflected. The situation that the static safety risk of the actual power grid occurs is often caused by the situations of sudden disconnection, severe change of new energy output and the like, and the change of the static safety of the system can be reflected by calculating the change situation of the first layer of evaluation index. The second-layer evaluation index is based on the node voltage out-of-limit risk index and the line power overload risk index obtained by the first-layer calculation, and the corresponding risk change index is calculated by the following formula:

wherein,representation->Node voltage out-of-limit risk change indexes at moment; />Representation->And (5) a line power overload risk change index at the moment.

(c) Third layer evaluation index

By introducing the comprehensive risk index as a third-layer evaluation index, the comprehensive risk condition of the power system can be reflected. The comprehensive risk index can be calculated by the following formula:

wherein,the weight corresponding to the node voltage out-of-limit risk change index is given; />The weight corresponding to the line power overload risk change index is given; / >And->The values of (2) may be set as desired, for example, all set to 1.

The integrated risk indicator is typically located in a relatively small fluctuation range when the power system is in a safe operation range. When the power system faces larger fluctuation (for example, a part of lines have larger failure probability due to extreme weather factors and the new energy output has severe fluctuation), the static safety risk of the power system is higher, and the calculated comprehensive risk index can generate relatively larger fluctuation.

(2) Building element fault probability model

The component fault probability model is used for quantifying the influence of target conditions (such as meteorological conditions) on different component fault rates in the operation process of the power system, can be derived according to the vulnerability and accident analysis of different components, and is generally described as follows:

wherein,is positioned at +.>The elements at->To->Probability of failure occurring within a time period; />Is->Time target variable (e.g. meteorological variable) +.>The value is +.>The position is->A conditional probability distribution of failure of the component at the location; />Is a dangerous event->Time->When it happens, the position is +.>Target atVariable->The value is +.>Probability distribution functions of (a) are provided.

Because of limitations of the power system measurement devices and communication requirements, only measurement data can be obtained at intervals, it is assumed that the target conditions in the two measurement periods are unchanged, and it is possible to obtain:

wherein,indicating time->To->Time Meteorological variable target variable->The value is +.>The position is->The conditional probability distribution of the failure of the element at that location can be fitted using a polynomial. />Is a dangerous event->Time to->Time->Persisting at +.>Target variable->The value is +.>Probability distribution functions of (a) are provided.

It will be appreciated that the element failure probability model input is real-time measurement data of each element, and the output is element failure probability.

And step S920, taking the risk index, the element failure rate and the power load data as input data, and predicting by adopting a hybrid neural network to obtain power operation data and risk level.

(1) And (3) data acquisition: based on the unified information model, the knowledge graph database and the data virtualization tool described in the above embodiments, a virtual table suitable for static stability analysis is generated. As described in the above embodiments, the data of all different service systems have been subjected to unified information modeling and multi-source data fusion, and the virtual table in this embodiment is the data that can be directly applied to static stability analysis.

(2) Feature engineering and selection: the power load data required for static stability analysis is extracted from the generated virtual table, and these features can be used to describe the state and stability of the power system, such as node voltage, generator power, load change rate, line parameters, etc.

(3) Static stability analysis and prediction: and carrying out static stability analysis and prediction on the extracted real-time data by using the trained mixed neural network. The input data can be risk indexes output by any one or more layers of the three layers of evaluation indexes, element fault probability output by an element probability model, power load data, new energy prediction data and the like, and the output data is stability indexes and risk grade ratings of the power system. The stability index may be, but is not limited to, node voltage, transformer load rate, rotor angle stability, frequency stability, line load rate, etc., to evaluate the stability status and potential risk of the power system.

In one embodiment, a hybrid neural network may be constructed using LSTM (long short term memory artificial neural network) and a fully connected neural network. In a specific implementation, input data can be input to an LSTM, so that the LSTM mines information such as history information, output node voltage, line load rate and the like, and then information output by the LSTM is input to a fully connected neural network, so that the fully connected neural network corrects the output of the LSTM to obtain output data.

In the training stage of the hybrid neural network, the hybrid neural network learns a great amount of historical data information in the power system. After training is completed, the trained hybrid neural network may be validated and evaluated using the validation dataset, e.g., the hybrid neural network may be evaluated using evaluation metrics such as accuracy, recall, F1 values, etc., to check the accuracy and generalization ability of the hybrid neural network.

In yet another embodiment, in the conventional technology, when the regulation service is required, a scheduling instruction is generally issued by a scheduling center. In this embodiment, a pure data driving-based analysis system is adopted, a plurality of edge clusters can autonomously formulate a scheduling policy, and a scheduling center integrates and coordinates to realize a distributed control service, and in this embodiment, a service processing request is a distributed control request. Referring to fig. 9, this can be achieved by:

(1) The edge cluster receives a distributed regulation request initiated at a business layer, analyzes a pre-generated knowledge graph index library, and acquires target power data of the power system, wherein the target power data comprises operation information of the power system. The operation information of the power system comprises historical operation information and/or real-time operation information, the operation information can be measurement data of each regional power grid, and the measurement data comprises data information such as current, voltage, active power, reactive power, switching state and the like related to each node/equipment;

(2) And sending the operation information to a cloud system (cloud scheduling center in fig. 9) through a cloud side cooperative interaction module, adjusting the operation information by the cloud scheduling center according to a preset rule or according to a trained whole network cooperative model, and returning the adjusted operation information to the edge cluster.

(3) An edge computing platform is deployed in a land edge cluster dispatching center, the edge computing platform inputs the adjusted operation information into a pre-deployed whole network model, a dispatching decision is predicted to be obtained, the dispatching decision comprises equipment or a regulating main body which needs to be regulated and controlled, and then a dispatching instruction is generated.

The full-network model may employ a neural network model, or other machine learning model that may generate scheduling decisions. The full-network model suitable for the intelligent agent can be selected from the model algorithm library in a self-adaptive mode according to the characteristics of each edge cluster or the intelligent agent. In one example, under four typical operation modes of traditional large, small, big and small, extreme weather and extreme conditions are considered, and the method is expanded into 12 operation scenes including typical scenes and extreme scenes. The model algorithm library comprises 12 typical intelligent decision models such as a bidirectional long and short time memory network (BLSTM), a Decision Tree (DTR), a Random Forest (RF) and the like; aiming at each specific decision scene, through training and iterative verification of historical data, screening out the whole network model under different scenes, and establishing a scene-model mapping relation to realize the self-adaptive matching of each edge cluster scheduling model.

The cloud computing center module is deployed in the cloud system, the cloud computing center module can be used in advance to train the full-network model according to the historical operation information of the power system, a trained full-network model is obtained, and the trained full-network model is deployed in the edge cluster.

Furthermore, the cloud computing center module can also periodically acquire the operation information of the power system, and update the model parameters of each full-network model by adopting the acquired operation information, so that the accuracy of the full-network model is ensured.

(4) And issuing the scheduling instruction to equipment or a regulating main body needing regulating through the edge gateway.

It will be appreciated that the above examples are all purely data-driven power system services based on the framework of the analysis system shown in fig. 1, and in practical applications, there are very many services that need to be processed, but the implementation principles are all communicated, and the present invention is not specifically exemplified.

It should be understood that, although the steps in the flowcharts related to the embodiments described above are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in the flowcharts described in the above embodiments may include a plurality of steps or a plurality of stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily performed sequentially, but may be performed alternately or alternately with at least some of the other steps or stages.

Based on the same inventive concept, the embodiment of the application also provides a pure data-driven power system analysis device for realizing the above-mentioned pure data-driven power system analysis method. The implementation of the solution provided by the device is similar to the implementation described in the above method, so the specific limitations in the embodiments of the power system analysis device or devices provided below may be referred to the limitations of the power system analysis method hereinabove, and will not be repeated herein.

In one embodiment, as shown in fig. 10, there is provided a pure data driven power system analysis device 1000 comprising: query information parsing module 1002, data query module 1004, data virtualization module 1006, and business processing module 1008, wherein:

a query information parsing module 1002, configured to determine data query information corresponding to a service processing request in response to the service processing request of the service end;

the data query module 1004 is configured to obtain target power data according to data query information and a pre-constructed knowledge graph index library, where the knowledge graph index library is a graph index library generated based on a pre-constructed unified information model, and the unified information model is a model generated by expanding based on a public information model;

A data virtualization module 1006, configured to map target power data to a virtual table, where the virtual table is a table that can be directly used when performing service processing;

and the service processing module 1008 is configured to perform service processing based on the virtual table, so as to obtain a service processing result.

In one embodiment, the apparatus 1000 further comprises: the similarity analysis module is used for determining the similarity between the first equipment class and the second equipment class existing in the public information model under the condition that the first equipment class which does not meet modeling requirements exists in the power data through analysis; the model expansion module is used for generating an expansion model according to the similarity, the first equipment class and attribute information corresponding to the first equipment class; and the model generation module is used for generating a unified information model according to the extension model and the public information model.

In one embodiment, the model extension module is configured to, when it is determined that a second device class similar to the first device class exists in the public information model based on the similarity, create a new class according to the similar second device class, construct an inheritance relationship between the new class and the similar second device class, and generate an extension model based on attribute information corresponding to the new class and the new class; when it is determined that a second equipment class similar to the first equipment class does not exist in the public information model based on the similarity, a new class is created according to the first equipment class, and an expansion model is generated based on the new class and attribute information corresponding to the new class.

In one embodiment, the data query module 1004 is configured to determine a target element corresponding to the data query information; and generating a query statement according to the target element, executing the query statement, and querying from a pre-constructed knowledge graph index library to obtain power data corresponding to the target element as target power data.

In one embodiment, the business process request is a risk assessment request, and the virtual form includes power load data required to process the risk assessment request; the apparatus 1000 further comprises: the calculation module is used for directly using the power load data in the virtual table to respectively generate a risk index and an element failure rate of the power system; and the model application module is used for taking the risk index, the element failure rate and the power load data as input data and adopting the hybrid neural network prediction to obtain power operation data and risk grade.

The above-described individual modules in the pure data driven power system analysis device may be implemented in whole or in part by software, hardware, and combinations thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.

In one embodiment, a computer device is provided, which may be a server, and the internal structure of which may be as shown in fig. 11. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is for storing power data. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program, when executed by the processor, implements a pure data driven power system analysis method.

It will be appreciated by those skilled in the art that the structure shown in fig. 11 is merely a block diagram of a portion of the structure associated with the present application and is not limiting of the computer device to which the present application applies, and that a particular computer device may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.

In an embodiment, a computer device is provided, comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the method according to any of the embodiments described above when executing the computer program.

In one embodiment, a computer readable storage medium is provided, on which a computer program is stored which, when executed by a processor, implements the method of any of the embodiments described above.

In an embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the method according to any of the embodiments described above.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the various embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magnetic random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (Phase Change Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as Static Random access memory (Static Random access memory AccessMemory, SRAM) or dynamic Random access memory (Dynamic Random Access Memory, DRAM), and the like. The databases referred to in the various embodiments provided herein may include at least one of relational databases and non-relational databases. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic units, quantum computing-based data processing logic units, etc., without being limited thereto.

The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The above examples only represent a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the present application. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application shall be subject to the appended claims.

Claims (8)

1. A method of pure data driven power system analysis, the method comprising:

responding to a service processing request of a service end, determining data query information corresponding to the service processing request, wherein the service processing request is a risk assessment request;

obtaining target power data according to the data query information and a pre-constructed knowledge graph index library, wherein the knowledge graph index library is an index library generated based on a pre-constructed unified information model, and the unified information model is a model generated by expanding based on a public information model;

Mapping the target power data to a virtual table, wherein the virtual table is a table which can be directly used when business processing is carried out, and comprises power load data required by processing the risk assessment request;

directly using the power load data in the virtual table to generate a risk index of the power system, wherein the power load data comprises node voltage amplitude values of all nodes which are actually measured, the risk index of the power system comprises a node voltage out-of-limit risk index and a line power overload risk index of the whole power system which are calculated based on the node voltage amplitude values, and the node voltage out-of-limit risk index of the whole power system is calculated by the following formula:

；

；

wherein,representing node voltage out-of-limit risk indexes of the whole power system; />Indicate->The per unit value of the node voltage amplitude of each node; />Indicate->Severity of individual node voltage violations;

the line power overload risk index of the whole power system is calculated by the following formula:

；

wherein,a line power overload risk indicator representing the entire power system; />Indicate->Load rate of the lines; / >Indicate->Severity of line overload;

establishing an element fault probability model, and taking the output of the element fault probability model as element fault probability, wherein the element fault probability model has the following formula:

；

wherein,is positioned at +.>The elements at->To->Probability of failure occurring within a time period; />Is->The weather variable value at the moment is +.>The position is->A conditional probability distribution of failure of the component at the location; />Is a dangerous event->At->The position at the moment of time is +.>Meteorological variable value at the site is +.>Probability distribution functions of (2);

taking the risk index, the element fault probability and the power load data of the power system as input data, and predicting by adopting a hybrid neural network to obtain power operation data and risk level, wherein the hybrid neural network is constructed by utilizing an LSTM and a fully connected neural network;

performing service processing based on the virtual form to obtain a service processing result;

the generation mode of the unified information model comprises the following steps:

under the condition that the power data are analyzed and the first equipment class which does not meet modeling requirements exists, similarity between the first equipment class and the second equipment class existing in the public information model is determined;

Generating an expansion model according to the similarity, the first equipment class and attribute information corresponding to the first equipment class;

and generating the unified information model according to the extension model and the public information model.

2. The method of claim 1, wherein the generating an extended model from the similarity, the first device class, and attribute information corresponding to the first device class comprises:

when a second equipment class similar to the first equipment class exists in the public information model based on the similarity, creating a new class according to the similar second equipment class, constructing an inheritance relationship between the new class and the similar second equipment class, and generating the expansion model based on the new class and attribute information corresponding to the new class;

when it is determined that a second equipment class similar to the first equipment class does not exist in the public information model based on the similarity, a new class is created according to the first equipment class, and the expansion model is generated based on the new class and attribute information corresponding to the new class.

3. The method according to claim 1, wherein the obtaining the target power data according to the data query information and the pre-constructed knowledge-graph index library includes:

Determining a target element corresponding to the data query information;

and generating a query statement according to the target element, executing the query statement, and querying from the pre-constructed knowledge graph index library to obtain power data corresponding to the target element as the target power data.

4. The method according to any one of claims 1 to 3, wherein the predicting the power operation data and the risk level using the hybrid neural network includes:

inputting the input data into an LSTM to obtain information output by the LSTM;

and inputting the information output by the LSTM into the fully-connected neural network to obtain output data of the fully-connected neural network, and taking the output data as the power operation data and the risk level.

5. A pure data driven power system analysis system, the system comprising:

the edge gateway is connected with the equipment object and is used for collecting original power data of the equipment object;

the edge cluster is in communication connection with the edge gateway and is used for responding to a service processing request of a service end, determining data query information corresponding to the service processing request, wherein the service processing request is a risk assessment request; obtaining target power data according to the data query information and a pre-constructed knowledge graph index library; mapping the target power data to a virtual table, the virtual table including power load data required to process the risk assessment request; directly using the power load data in the virtual table to generate a risk index of the power system, wherein the power load data comprises node voltage amplitude values of all nodes which are actually measured, the risk index of the power system comprises a node voltage out-of-limit risk index and a line power overload risk index of the whole power system which are calculated based on the node voltage amplitude values, and the node voltage out-of-limit risk index of the whole power system is calculated by the following formula: Wherein->Representing node voltage out-of-limit risk indexes of the whole power system; />Indicate->The per unit value of the node voltage amplitude of each node; />Indicate->Severity of individual node voltage violations; the line power overload risk index of the whole power system is calculated by the following formula: wherein->A line power overload risk indicator representing the entire power system;indicate->Load rate of the lines; />Indicate->Severity of line overload; establishing an element fault probability model, and taking the output of the element fault probability model as element fault probability, wherein the element fault probability model has the following formula:wherein->Is positioned at +.>The elements at->To->Probability of failure occurring within a time period; />Is->The weather variable value at the moment is +.>The position is->A conditional probability distribution of failure of the component at the location; />Is a dangerous event->At->The position at the moment of time is +.>Meteorological variable value at the site is +.>Probability distribution functions of (2); taking the risk index, the element fault probability and the power load data of the power system as input data, and predicting by adopting a hybrid neural network to obtain power operation data and risk level, wherein the hybrid neural network The network is constructed by utilizing LSTM and a fully-connected neural network; performing service processing based on the virtual form to obtain a service processing result; the knowledge graph index library is an index library generated based on a pre-constructed unified information model, the unified information model is a model generated by expanding based on a public information model, and the virtual table is a table which can be directly used when the edge cluster performs service processing; the generation mode of the unified information model comprises the following steps: under the condition that the power data are analyzed and the first equipment class which does not meet modeling requirements exists, similarity between the first equipment class and the second equipment class existing in the public information model is determined; generating an expansion model according to the similarity, the first equipment class and attribute information corresponding to the first equipment class; generating the unified information model according to the extension model and the public information model;

and the cloud system is in communication connection with the edge cluster and is used for receiving the service processing result sent by the edge cluster and performing operation control based on the service processing result.

6. A pure data driven power system analysis device, the device comprising:

the query information analysis module is used for responding to a service processing request of a service end, determining data query information corresponding to the service processing request, wherein the service processing request is a risk assessment request;

the data query module is used for obtaining target power data according to the data query information and a pre-built knowledge graph index library, wherein the knowledge graph index library is an index library generated based on a pre-built unified information model, and the unified information model is a model generated by expansion based on a public information model;

the data virtualization module is used for mapping the target power data to a virtual table, wherein the virtual table is a table which can be directly used when business processing is carried out, and comprises power load data required by processing the risk assessment request;

the calculation module is configured to directly use the power load data in the virtual table to generate a risk indicator of the power system, where the power load data includes node voltage magnitudes of all nodes that are actually measured, the risk indicator of the power system includes a node voltage out-of-limit risk indicator and a line power overload risk indicator of the whole power system that are calculated based on the node voltage magnitudes, and the node voltage out-of-limit risk indicator of the whole power system is calculated by the following formula: Wherein->Representing node voltage out-of-limit risk indexes of the whole power system; />Indicate->The per unit value of the node voltage amplitude of each node; />Indicate->Severity of individual node voltage violations; the line power overload risk index of the whole power system is calculated by the following formula: /> Wherein->A line power overload risk indicator representing the entire power system; />Indicate->Load rate of the lines; />Indicate->Severity of line overload; establishing an element fault probability model, and taking the output of the element fault probability model as element fault probability, wherein the element fault probability model has the following formula: />Wherein->Is positioned at +.>The elements at->To->Probability of failure occurring within a time period; />Is->The weather variable value at the moment is +.>The position is->A conditional probability distribution of failure of the component at the location; />Is a dangerous event->At->The position at the moment of time is +.>Meteorological variable value at the site is +.>Probability distribution functions of (2);

the model application module is used for taking the risk index, the element fault probability and the power load data of the power system as input data, and predicting by adopting a hybrid neural network to obtain power operation data and risk level, wherein the hybrid neural network is constructed by utilizing an LSTM and a fully-connected neural network;

The service processing module is used for carrying out service processing based on the virtual form to obtain a service processing result;

the similarity analysis module is used for determining the similarity between the first equipment class and the second equipment class existing in the public information model under the condition that the first equipment class which does not meet modeling requirements exists in the power data through analysis;

the model expansion module is used for generating an expansion model according to the similarity, the first equipment class and attribute information corresponding to the first equipment class;

and the model generation module is used for generating the unified information model according to the extension model and the public information model.

7. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 4 when the computer program is executed.

8. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 4.