CN112102111A

CN112102111A - Intelligent processing system for power plant data

Info

Publication number: CN112102111A
Application number: CN202011029005.9A
Authority: CN
Inventors: 刘自愿; 林平; 谢鹏; 唐亮
Original assignee: Huadianfu New Guangzhou Energy Co ltd
Current assignee: Huadianfu New Guangzhou Energy Co ltd
Priority date: 2020-09-27
Filing date: 2020-09-27
Publication date: 2020-12-18
Anticipated expiration: 2040-09-27
Also published as: CN112102111B

Abstract

The invention provides an intelligent processing system for power plant data, which comprises: an operation and maintenance management module; a security management module; a data acquisition and extraction module; a data storage module; a general analysis and calculation module; intelligent calculation and analysis module: based on a machine learning algorithm, an intelligent algorithm model is established and trained by utilizing distributed computing resources, and a prediction model is established through data after multidimensional associated data analysis; an edge calculation module; the microservice issuing module: the system is used for issuing and managing system bearing data, computing tasks and external services; and a data management module. This power plant data intelligence processing system carries out the degree of depth through fusing with technologies such as big data, thing networking, cloud calculation, artificial intelligence, can realize the accurate collection of each system data of power plant, realizes the intelligent processing of each system data, improves power generating equipment automation, digitization, visual and intelligent degree and equipment reliability and the availability ratio.

Description

Intelligent processing system for power plant data

Technical Field

The invention relates to the technical field of electric power data processing, in particular to an intelligent processing system for power plant data.

Background

Along with the implementation of national energy policies, energy conservation, consumption reduction and operation and maintenance cost reduction are essential requirements for improving operation benefits of thermal power plants. In most of domestic power plants at present, various monitoring systems are independently administrative, such as a thermal control system, an auxiliary control system, an electric automation system, a booster station control system (NCS) and the like, and meanwhile, a user needs to face different monitoring systems, different system providers, different spare parts and the like, so that the maintenance workload of the power plant is greatly increased. Because each control system is independent, the data processing of each control system is also different, so that the data resources of the whole plant cannot be shared, and various high-level applications need to face different data processing forms, thereby bringing more difficulties to the high-level applications, data analysis and fault analysis.

In the prior art, a power plant is also generally built with a data processing system, but the existing data processing system mainly has the following defects:

1) the system lacks a stable operation and maintenance management and security management system, the operation stability and maintainability of the system are poor, the storage and calculation service of a user and the control capability of storage resources and calculation resources are weak, and the system is easy to lose or disorder data and have poor traceability performance after being attacked from the outside;

2) the data acquisition is incomplete, the efficiency is low, and the safety and the stability of the data acquisition cannot be guaranteed;

3) the data storage architecture is complex and lack of flexibility, the service is tightly coupled with the infrastructure, the application system is limited by the dependence relationship between software and hardware, and meanwhile, the traditional data center is complex in operation and maintenance management, low in resource utilization rate and high in maintenance cost;

4) the data processing capability is weak, the intelligent learning and modeling capability of the algorithm is not realized, and the functions of index analysis, equipment diagnosis, fault early warning and the like cannot be realized;

5) the system does not have a data management system, cannot help an electric power enterprise to standardize a data flow, manage main data in the enterprise, cannot improve the data quality of the enterprise, and ensures that the enterprise obtains accurate, timely and complete data support in business operation management.

With the rise of new technologies such as big data, internet of things, cloud computing and artificial intelligence, how to realize the deep fusion of the technologies and the power plant services and realize the intelligent processing of each system data, the automation, digitization, visualization and intelligence degree of power generation equipment, the reliability and availability of the equipment are improved, the workload and working strength of operators are reduced, the safety precaution level is strengthened, the production management and cost control modes are innovated, the operation benefit of a unit is improved, the competitive power of the power market is participated, and the technology problem which is urgently needed to be solved in the power industry is achieved.

Disclosure of Invention

Aiming at the defects of the prior art, the invention provides the intelligent processing system for the power plant data, which can realize accurate acquisition of each system data of the power plant, realize intelligent processing of each system data, and improve the automation, digitization, visualization and intelligence degree of power generation equipment, the reliability and availability of the equipment by carrying out deep fusion with technologies such as big data, Internet of things, cloud computing, artificial intelligence and the like.

In order to realize the technical scheme, the invention provides an intelligent processing system for power plant data, which comprises: the operation and maintenance management module: the system is used for realizing system physical node management, distributed component management, multi-tenant and user management, data storage and job management; a safety management module: the system is used for data management of physical security, component security, network security and application software security, forms a multi-dimensional and three-dimensional security support, and guarantees the stable operation of the system and the data security; a data acquisition and extraction module: the acquisition of the heterogeneous data in the target area is realized through protocol acquisition software based on IEC60870 and an ETL extraction tool; a data storage module: establishing an operation data storage center, a time sequence data storage center, a data warehouse and unstructured data storage based on an Hadoop ecological open source component; general analysis and calculation module: an algorithm model base is established, and real-time, quasi-real-time and off-line calculation is realized by relying on the distributed calculation capacity of Storm, Spark and MapReduce; intelligent calculation and analysis module: based on a machine learning algorithm, an intelligent algorithm model is established and trained by utilizing distributed computing resources, and a prediction model is established through data after multidimensional associated data analysis; an edge calculation module: the prediction model becomes an important calculation service node of a big data system by using the calculation capacity of the intelligent equipment, and then the intelligent terminal can perform real-time analysis and diagnosis according to the obtained model by issuing an algorithm model through the system; the microservice issuing module: the system is used for issuing and managing system bearing data, computing tasks and external services; the data management module: the method comprises the steps of managing main data and metadata, uniformly managing various heterogeneous data, establishing a management system and providing support for application of an upper-layer service system.

Preferably, the operation and maintenance management module is a unified visual management and control module developed and constructed based on open source components, and is open to developers, operation and maintenance personnel and managers, so that unified visual management and control of data, applications and resources are realized, and the method specifically includes: infrastructure and component management module: the system is used for installing distributed clusters of Hadoop, Yann, Zookeeper, Spark, HDFS, HBase, Hive, Storm and MapReduce components, realizes the installation and management of the components and the components depending on the components, supports the automatic installation and deployment of the components, uses tools for automatic installation, monitors the operation condition of a big data cluster, including the state and the operation load of a host node, a network node and a frame, simultaneously monitors the occupation condition of software and hardware of each node of the system, and if an abnormal condition occurs to a monitored object, the monitoring system can send an alarm notice on a related management alarm page; a multi-tenant management module: the method is used for sharing the same system or program in a multi-user environment, and ensuring the isolation of data among users, so that a big data system is fully utilized, service is provided for more users, the data safety of the users and the normal operation of application programs are ensured, and the traceability is realized while resource sharing is realized; resource and job management module: the method has the advantages that the storage resources are distributed, a visual storage resource distribution interface and a visual storage resource retrieval interface are provided, the computing resources are distributed by the resource scheduling component according to the computing condition, a visual job submission function and a plurality of job arrangement functions are provided, the jobs are conveniently issued and the process is conveniently arranged, meanwhile, a job monitoring interface is also provided, the operation running state and the resource using condition are conveniently known, and the job running log is conveniently checked; a log management module: combining Logstash, elastic search and Kibana tools to construct a centralized log management system, collecting logs of different components, storing the logs in a centralized manner, and providing a full-text retrieval function; an alarm management module: the system has a configurable alarm management function and can configure alarms of different levels according to alarm rules; an alarm notification mechanism and an alarm manager can be configured according to actual service requirements.

Preferably, the security management module comprises a system security management module and a physical security management module, wherein the system security management module manages from a host layer, a component layer, a network layer and an interface API layer, the security authentication of the host layer is realized by identity authentication, access control, security audit, intrusion prevention, malicious code prevention and resource control, the security authentication of the component layer is realized by a Ranger framework, the security authentication of the network layer is realized by a Kerberos protocol, and the security authentication of the interface API layer is realized by user login and JWT authentication; the physical security management module manages from physical location selection, physical access control, theft and damage prevention, lightning and fire prevention, water and moisture prevention, static electricity prevention, temperature and humidity control, power supply, and electromagnetic protection.

Preferably, the data acquisition and extraction module is used for acquiring structured, semi-structured and unstructured data generated in the production operation and operation management process of the power plant, acquiring production real-time data from each control system according to IEC61850 and IEC61970 standards, extracting discrete historical data from the subsystems through an ETL extraction tool, and dividing the discrete historical data according to data types, wherein the acquired data mainly comprises production time sequence data, document type data and discrete relational data extracted from relational data of other systems of the power plant.

Preferably, the data storage module includes: the system comprises a time sequence data storage module, a structured data storage module and an unstructured data storage module, wherein the time sequence data storage module is used for storing real-time data which are generated in the production process and stored on the basis of time sequences; the structured data storage module is used for storing operation records and management logs generated in the production or management process of the power plant subsystem; the unstructured data storage module is used for storing information which is irregular and cannot be expressed by a two-dimensional logic table, such as audio, images and HTML, and the unstructured data is stored by adopting a NoSQL database.

Preferably, the general analysis and calculation module comprises: the system comprises a real-time calculation module, a historical statistics calculation module and a state calculation module, wherein the real-time calculation module is used for calculating real-time indexes including alarm point statistics and real-time integral electric quantity based on Storm distribution; the historical statistic calculation module is used for calculating indexes for historical data statistics; the state calculation module is used for calculating the fault information of the unit, giving an alarm and displaying in real time, and can store the information of data generation, end and grade.

Preferably, the intelligent calculation and analysis module establishes a unit performance model, a thermodynamic system model and an equipment state model according to real-time and historical data, establishes an early warning model according to a monitoring range and an object, associates measuring points, performs rule definition and model training, automatically obtains a residual error initial value and other important default setting parameters through data training after the model is established, can conveniently change the model parameters at any time, records the trace of parameter modification in the system, obtains variables which cannot be directly obtained through the measuring points, can take expected values, residual errors and residual error alarms in the original measuring points, the calculation measuring points and the prediction measuring points as input values of the calculation measuring points, the calculation measuring points are positioned under a project and can be quoted by the models and rules under the project, the configuration function of the calculation measuring points is arranged under the project management, and the configuration of the models, The rule configurations are located at the same level.

Preferably, the edge calculation module includes: the system comprises a measurement data display and analysis module, a performance online analysis module, a performance offline analysis module and an operation and maintenance decision support module, wherein the measurement data display and analysis module is used for comparing data of the same measuring point at different moments, calculating absolute difference and relative difference of numerical values at different moments, marking differences of different degrees by using colors, displaying time sequence data of the measuring points by using a trend graph, the abscissa of the trend graph is time, the ordinate of the trend graph is measuring point numerical value, a plurality of measuring point data are displayed in the trend graph, different measuring points are distinguished by using colors, a continuous curve in a period of time is drawn for a certain measuring point, and the change of a time interval is supported; the performance online analysis module utilizes the comparison table to compare and analyze different data types of the same measuring point, the data types comprise a measured value, a verification value and an expected value, the comparison result comprises an absolute deviation and a relative deviation, and a large deviation can be identified by colors; the performance offline analysis module is used for supporting a user to change input data; the operation and maintenance decision support module is used for providing an auxiliary decision for the operation of the power plant by using the simulator on the basis of index calculation and operation deviation analysis, and can generate an economic report in a customized manner through historical data simulation.

Preferably, the micro-service publishing module adopts a micro-service technical architecture, each service has a processing and lightweight communication mechanism, and can be deployed on a single or multiple servers, an API gateway is used as a unique entrance of all clients, receives client requests and routes the client requests to a proper service, the API gateway opens different APIs according to different clients, access control load balancing is realized through the API gateway, security verification can also be realized whether the clients have authority to access the service, different service services are mutually called through REST, the service is deployed in a container, and the system provides registration and discovery, heartbeat monitoring, current limiting, degradation and fusing of the service.

Preferably, the main data management in the data management module is to coordinate and manage basic data related to core business entities of the power plant by adopting a series of rules, applications and technologies, so as to provide high-quality main data which is applied across businesses, is consistent in use and is shared for the power plant, and reduce the cost and the complexity; metadata management is used to describe concepts, relationships and rules related to technology, business and management in a system, including definitions of objects and data structures within the system, mapping of source data to destination data, description of data transformation, metrics, personnel roles, job responsibilities and management processes.

The intelligent processing system for the power plant data provided by the invention has the beneficial effects that:

1) this power plant data intelligent processing system builds based on open source Hadoop ecological component, and the use includes: components such as Hadoop, Yarn, Zookeeper, Spark, Storm, HBase, Hive, Kafka and the like can realize real-time data acquisition, historical (discrete) data extraction, data governance, data preprocessing, data storage (structured and unstructured), data analysis and calculation and service release by adopting an SOA service framework design based on the distributed components, break through a one-to-one traditional mode, and efficiently utilize a physical server through a resource allocation and scheduling mechanism at the bottom layer, so that the requirement of running multiple sets (multiple types) of operating systems on one server is met.

2) This power plant data intelligence processing system carries out the degree of depth through with technologies such as big data, thing networking, cloud computing, artificial intelligence and fuses, can realize the accurate collection of each system data of power plant, realizes the intelligent processing of each system data, improves power generating equipment automation, digitization, visual and intelligent degree and equipment reliability and the availability ratio.

3) This power plant data intelligent processing system can improve the operating stability and the maintainability of system through setting up dedicated fortune dimension management module, improves user's storage, computational service and to storage resource, computational resource's management and control ability, through setting up dedicated safety management module, has solved the system because suffer outside attack easily and lead to data to lose or in disorder, problem that traceability can be poor.

4) This power plant data intelligence processing system has solved the data acquisition incomplete through the data acquisition and the extraction module that use IEC 60870-based protocol collection software and ETL extraction instrument, and data acquisition is not comprehensive, and real-time data acquisition's accuracy is poor, especially discrete relation type data acquisition is incomplete, the inefficiency problem.

5) The intelligent power plant data processing system is based on Hadoop ecological open source components, an operation data storage center, a time sequence data storage center, a data warehouse and unstructured data storage are established, a data storage framework is simplified, the flexibility of data storage is improved, the utilization rate of data storage resources is improved, and the maintenance cost is reduced.

6) The intelligent processing system for the power plant data can build and train an intelligent algorithm model by utilizing distributed computing resources based on a machine learning algorithm through building a general analysis and calculation module, an intelligent calculation and analysis module and an edge calculation module, can build a prediction model through data after multidimensional associated data analysis, enhances the intelligent processing capability of the data, and can realize index analysis, equipment diagnosis and fault early warning.

7) This power plant data intelligence processing system possesses the data governance system, can help the electric power enterprise to standardize the data flow, and the inside main data of management enterprise promotes enterprise data quality, guarantees that the enterprise obtains accuracy, timely and complete data support in the business operation management.

Drawings

FIG. 1 is a SOA service framework diagram of the present invention.

Fig. 2 is an architecture diagram of a security management module according to the present invention.

FIG. 3 is an architecture diagram of a data acquisition and extraction module of the present invention.

FIG. 4 is a diagram of the architecture of a data storage module according to the present invention.

FIG. 5 is a block diagram of a microservice publishing module of the present invention.

FIG. 6 is a block diagram of a data administration module according to the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments obtained by a person skilled in the art without making any inventive step are within the scope of the present invention.

Example (b): an intelligent processing system for power plant data.

Referring to fig. 1 to 6, an intelligent processing system for power plant data is designed by using an SOA service framework based on distributed components (as shown in fig. 1), and from the viewpoint of data flow direction, the intelligent processing system includes main real-time data acquisition, historical (discrete) data extraction, data governance, data preprocessing, data storage (structured and unstructured), data analysis and calculation, and service distribution, and in order to ensure stable operation and maintainability of the system, two support means of operation and maintenance management and safety management are introduced, which specifically includes:

the operation and maintenance management module comprises: the system is used for realizing system physical node management, distributed component management, multi-tenant and user management, data storage and job management; operation and maintenance management is mainly the management of large data system infrastructure and distributed components running on the large data system infrastructure, and also comprises the management of users and multiple tenants. The operation and maintenance management provides storage and computing services for users and management and control services for storage resources and computing resources. The management and control service is based on the unified visual management and control ability of open source subassembly development construction, mainly faces to development personnel, operation and maintenance personnel, managers, realizes the unified visual management and control of data, application, resource, and it includes: multi-tenant management, resource management, job management, log management, alarm management, and the like.

The operation and maintenance management module is a unified visual management and control module constructed based on open source component development, is open to developers, operation and maintenance personnel and managers, realizes unified visual management and control of data, application and resources, and specifically comprises:

11) infrastructure and component management module: the method is used for installing the distributed cluster of the Hadoop, Yarn, Zookeeper, Spark, HDFS, HBase, Hive, Storm and MapReduce components, realizes the installation and management of the components and the components depending on the components, supports the automatic installation and deployment of the components, and uses tools to carry out automatic installation. Monitoring of the infrastructure is mainly to view the operation conditions of the big data cluster from the whole, including the states and operation loads of the host nodes, the network nodes and the racks. And simultaneously, the occupation conditions of software and hardware of each node of the monitoring system, such as resources of a CPU (central processing unit), a hard disk, a memory and the like, are monitored, and if the monitored object has an abnormal condition, the monitoring system sends an alarm notice on a related management alarm page. The current component monitoring tools of big data systems mainly include Cloudera Manager, Ambari, and other domestic self-developed management systems. In view of the advantages of Ambari in version control, secondary development, the number of integrated components, openness and the like, the module develops an Ambari-based component management tool. Ambari is an open source system for deploying, managing and monitoring Hadoop cluster services, provides a cluster which is installed in a guide guiding mode, can install Hadoop services on any host, provides a configuration function of the Hadoop services, provides cluster management functions of starting, stopping and the like, provides an instrument panel for monitoring the health state of the Hadoop cluster, provides a set of health index system for collecting monitoring data, provides a set of early warning framework, and can realize notification and early warning by combining preset monitoring indexes. Ambari provides a set of strong management and maintenance functions for Hadoop service, including cluster users, service configuration, service monitoring, auxiliary tools and the like.

12) A multi-tenant management module: large data processing systems are resource and cost intensive information systems that provide powerful storage and computing power, and provide services to multiple users simultaneously. Multiple users use one system at the same time, and in order to ensure normal service, it is necessary to ensure that application programs among the users do not interfere with each other, and ensure that used resources and data are isolated from each other. Multi-tenant (Multi-tenacy), or Multi-Tenancy, is a software architecture technology that enables sharing of the same system or program in a Multi-user environment and still ensures isolation of data between users. The multi-tenant management of the big data processing system adopts a multi-tenant technology, so that on one hand, the big data processing system is fully utilized, and services are provided for more users; meanwhile, data safety of the user and normal operation of the application program are guaranteed. The tenants have the resources and data isolated from each other and do not influence each other, which can cause another problem that the data cannot be shared. In this case, for the same tenant, a plurality of members can be allocated, the members use the same tenant account, but have respective login user names and passwords, and the login and logout of the members are recorded on the case, so that the resource sharing is realized, and the traceability is also realized.

13) Resource and job management module: in the data processing system, under the environment of supporting multiple tenants, the allocation of resources is inevitable. Resource management can solve the problems of resource allocation and task viewing. The resource allocation mainly refers to allocation of storage resources, and provides a visual storage resource allocation interface and a storage resource retrieval interface. The computing resources are scheduled and distributed by the resource scheduling component according to the computing condition, and the data processing system provides a functional interface for retrieving the task execution condition according to the tenants. Due to the fact that a large number of distributed computing tasks are carried, the computing tasks based on distributed components such as Spark, Storm, MapReduce and the like need real-time monitoring and management. These computing tasks (programs) are submitted to the system for running in a job-like manner, and distributed computing is performed in the system. The system provides a visual job submitting function and a plurality of job arranging functions, and can conveniently release jobs and arrange flows. Meanwhile, an operation monitoring interface is provided, so that the operation state and the resource use condition of the operation can be conveniently known, and the operation log can be conveniently checked.

14) A log management module: the logs of the open source components have respective management modes and are dispersed in the cluster. If a problem occurs, a plurality of places are usually searched for final positioning, which is not beneficial to solving the problem. This problem is particularly acute in larger scale scenarios. In order to solve the problem caused by log dispersion, a set of centralized log management system needs to be constructed, logs of different components are collected and stored in a centralized manner, and a full-text retrieval function is provided. Logstash is a tool used to gather, analyze, and filter logs. It supports almost any type of log, including system logs, error logs, and custom application logs. It can receive logs from many sources including syslog, messaging (e.g., RabbitMQ), and JMX, which can output data in a variety of ways including email, websockets, and Elasticsearch. The Elasticissearch is a real-time full-text search and analysis engine and provides three functions of collecting, analyzing and storing data; the system is a set of structures such as open REST, JAVA API and the like, and provides an efficient search function and an extensible distributed system. It is built on the Apache Lucene search engine library. Kibana is a Web-based graphical interface for searching, analyzing and visualizing log data stored in the elastic search. It utilizes the REST interface of the Elasticsearch to retrieve data, allowing the user to customize the personalized view. The system combines the characteristics of Logstash, Elasticisearch and Kibana to construct a centralized log management system, and can well solve the problems of collection, retrieval, statistical analysis and the like of large-scale logs.

15) An alarm management module: alarm management is an important link for operation and maintenance monitoring, and is related to stable operation of the system. Because the system has many components and many alarms, a centralized alarm system is needed to solve the problems of alarm display, inquiry and the like. The alarm management has configurable alarm management capability, and can configure alarms of different levels according to alarm rules; an alarm notification mechanism, an alarm manager and the like can be configured according to actual service requirements.

(II) a safety management module: including system security, which is managed from the host layer, component layer, network layer (gateway), interface API layer, and physical security (see fig. 2 for details). The specific management is as follows:

21) a host layer: the security authentication of the host layer is realized by identity authentication, access control, security audit, intrusion prevention, malicious code prevention and resource control. For example, identity identification and authentication are performed on login users, accounts and permissions are allocated, a security audit function covering each user is started, possible bugs can be found, the bugs can be repaired in time after full test and evaluation, measures for protecting the integrity of important system programs or files are provided, recovery measures are taken when the integrity of the important system programs or files is detected to be damaged, the maximum use limit of a single user or process on system resources is limited, and the like.

22) Assembly layer: the security authentication of the component layer is implemented by a Ranger framework. Although HDFS, Hive, Hbase and the like have respective authority management functions, the HDFS, the Hive, the Hbase and the like are too distributed and have original configuration modes, management is not facilitated, and Ranger provides a uniform data authorization and management interface for a plurality of components in a Hadoop ecosystem. The method can perform fine-grained data access control on a plurality of components in the Hadoop ecosystem, such as HDFS, Hive, Hbase and the like. By operating the Range console, an administrator can easily control user access rights by configuring policies.

23) Network layer: the secure authentication at the network layer is achieved by the Kerberos protocol. The Kerberos protocol provides powerful communication encryption and authentication services between a server and a client application program through a powerful key system, and in a cluster authenticated by the Kerberos protocol, a client does not directly perform communication authentication with its server but completes mutual authentication through an independent service such as a kdc (key Distribution center). Meanwhile, Kerberos can also encrypt all communication between services to ensure privacy and integrity of the services.

24) Interface API layer: the security authentication of the interface API layer is achieved by user login and JWT authentication. The user login needs to be capable of identifying identity, a login authentication mechanism of the user needs to be added, and a third-party system needs to pass through a user login process before calling an interface API. JWT (JSONWeb Token) authentication is a Token string generated by a server, encrypted and encoded with base64, and the client needs to transmit the Token in the request header of each request, and the server can identify the validity of the user after the Token is taken.

Physical security is managed from physical location selection, physical access control, theft and vandalism prevention, lightning and fire prevention, water and moisture prevention, static electricity prevention, temperature and humidity control, power supply, electromagnetic protection, and the like.

25) Physical location selection: the machine room site is selected in a building with the capabilities of shock resistance, wind resistance, rain resistance and the like; the machine room floor is avoided in the top floor or basement of the building, otherwise the water and moisture proofing measures are enhanced.

26) Physical access control: an electronic access control system is configured at the entrance and exit of the machine room to control, identify and record the entering personnel; ensuring that an equipment room for storing, processing and analyzing the bearing big data is positioned in China; ensure that data is cleared or destroyed in China.

27) Theft and damage prevention: fixing the equipment or the main part, and arranging an obvious mark which is difficult to remove; laying the communication cable at a hidden place, wherein the communication cable can be laid underground or in a pipeline; an anti-theft alarm system of a machine room or a video monitoring system of a special person on duty is arranged.

28) Lightning protection and fire prevention: various cabinets, facilities, equipment and the like are safely grounded through a grounding system; measures are taken to prevent induction thunder, such as arranging a lightning protection protector or an overvoltage protection device. The automatic fire fighting system for the fire disaster is arranged, so that the fire condition can be automatically detected, the automatic alarm can be realized, and the automatic fire extinguishing can be realized; the machine room and the related working room and auxiliary room are made of fire-resistant building materials; the machine room is divided into areas, and isolation fire prevention measures are arranged between the areas.

29) Water and moisture proofing: measures are taken to prevent rainwater from permeating through windows, roofs and walls of the machine room; measures are taken to prevent the moisture condensation of the water vapor in the machine room and the transfer and permeation of the underground accumulated water; and a detection instrument or element sensitive to water is installed, and waterproof detection and alarm are carried out on the machine room.

210) And (3) antistatic: installing an anti-static floor and adopting necessary grounding anti-static measures; measures are taken to prevent the generation of static electricity, such as a static eliminator, wearing a static electricity prevention bracelet, and the like.

211) Temperature and humidity control: and automatic temperature and humidity adjusting facilities are arranged, so that the temperature and humidity of the machine room are changed within the allowable range of equipment operation.

212) Power supply: configuring a voltage stabilizer and overvoltage protection equipment on a power supply line of a machine room; providing a short-term backup power supply, at least meeting the normal operation requirement of the equipment under the condition of power failure; redundant or parallel power cabling is provided to power the computer system.

213) Electromagnetic protection: the power line and the communication cable are laid in an isolated way, so that mutual interference is avoided; electromagnetic shielding is applied to key equipment.

(III) a data acquisition and extraction module: the method is used for collecting structured, semi-structured and unstructured data generated in the production operation and operation management process of the power plant. Referring to IEC61850 and IEC61970 standards, real-time production data is collected from each control system, discrete historical data is extracted from subsystems through an ETL extraction tool, and the collection system framework is shown in fig. 3.

The data collected by the collecting system mainly comprises production time sequence data, document type data and discrete relational data extracted from relational data of other systems of the power plant.

31) Production of real-time data: the production real-time data is also called time sequence data, and is mainly a data sequence recorded by the same index according to time sequence. The acquisition system supports at least the following acquisition protocols: IEC101, IEC102, FFC102, IEC104, TCPModbus, COM Modbus, Ministry-issued CDT, OPC, etc. The real-time data acquisition interface adopts a multithreading technology and supports multi-path concurrent communication. The real-time data acquisition software has an edge calculation function and supports basic operators such as Summation (SUM), Averaging (AVG), extremum solving (MAX/MIN) and the like.

32) File type data: the ETL extraction tool can be used for realizing data acquisition, the file types which cannot be realized by the ETL tool can be independently developed to acquire interface programs, and acquisition protocols support FTP (file transfer protocol), SFTP (small form-factor pluggable) and SSH (secure Shell) modes.

33) Relational data: the relational data acquisition adopts an ETL extraction tool.

The ETL extraction tool provides the following functions:

a) connect to traditional relational databases and obtain data, including Orcale, SQL Server, IBM DB2, MySQL, PostgreSQL, and so on.

b) Data is obtained from an ASCII file with delimiters and a fixed format.

c) Data is obtained from the XML file.

d) Data is obtained from popular office software, such as Access databases and Excel spreadsheets.

e) And downloading data by adopting FTP, SFTP and SSH modes.

f) Data can also be obtained from Web Services or RSS.

g) Support concurrent and cluster deployment.

In order to ensure the safety and stability of data acquisition, the acquisition system meets the following requirements:

1) the operation systems of the acquisition interface machine are only limited to Unix, Debian, CentOS, congener and kylin.

2) The acquisition system has the functions of local caching and breakpoint continuous transmission.

3) And a supplementary function is supported for the electric quantity data.

4) The acquisition system adopts a friendly visual configuration interface.

5) The acquisition interface machine has the functions of interface and data state diagnosis and uploads the interface state to the data center.

(IV) a data storage module: and establishing an operation data storage center, a time sequence data storage center, a data warehouse and unstructured data storage based on the Hadoop ecological open source component.

The traditional data storage center is complex in structure and lack of flexibility, services are tightly coupled with infrastructure, an application system is limited by the dependence relationship between software and hardware, and meanwhile, the traditional data storage center is complex in operation and maintenance management, low in resource utilization rate and high in maintenance cost. The data storage center of the data storage module can be centered on users, is oriented to services, and is based on high-efficiency, green and software-defined IT and network infrastructure, so that higher performance, density, integration level and energy efficiency, modularization and componentization are realized. The abstraction, pooling, deployment and management of infrastructure resources in the whole data storage center are realized through the server virtualization and other technologies, and customized and differentiated application and service requirements are met.

In order to solve the problems of single-point faults, measurement point capacity limitation, data type diversification, index calculation efficiency and the like, the power plant data storage center framework is based on distributed components, a deployment scheme combining a traditional relational database and a Hadoop non-relational database is needed to meet the storage requirements of data type diversification, and the framework of a data storage module is shown in FIG. 4.

Data storage systems can be classified into time-series data storage, structured data storage, and unstructured data storage according to the type of data storage.

41) Time sequence data storage: the time series data is also called real-time data, and is mainly real-time data which is generated in the production process and stored on the basis of time series. To meet the field requirements, the time series data storage can meet the following requirements:

A. in order to meet the real-time requirement of a production management system, the analog quantity data acquisition frequency is changed and stored in the second level, and the switching value (a disconnecting link, a circuit breaker, a protection signal and the like) meets the change and storage in the millisecond level.

B. In order to meet the requirement of plant-level data analysis application, the data point has the capacity of not less than 3 ten thousand points and can support dynamic capacity expansion.

C. The database adopts a compression storage mode and supports lossless compression, so that the reality and the integrity of the data are ensured.

D. The database interface has high response speed, and the data read-write supports multi-point batch operation. In the local area network, the real-time data read-write time is less than 1.5s/1000 points. The historical stored value read takes <2 s/(stored value 10000, time span within 1 week).

42) And (3) structured data storage: the structured data mainly comprises operation records, management logs and the like generated in the production or management process of the power plant subsystem. Structured data is primarily stored in relational and non-relational databases. The structured data store mainly contains the following contents:

A. the data storage supports the storage and query mode of Key-Value, and can quickly locate query information.

B. Supports SQL query statement query, and comprises functions of storing procedures, triggering devices and the like.

C. The number of client connections (simultaneously online) supported by the database is no less than 25.

D. The database supports concurrent request requirements.

E. The data are coded uniformly, and the coding mode meets the requirements of the China Hua-Dy group material coding system.

43) Unstructured data storage: the unstructured data stores information that is irregular, and cannot be expressed in a two-dimensional logical table, mainly for audio, images, HTML, and the like. The unstructured data is stored by a NoSQL database, and the function and performance of the NoSQL database are as follows:

F. distributed deployment is supported, and nodes can be dynamically expanded.

G. And (4) designing by adopting a native Hadoop ecosystem component.

H. And the data of the object type is easy to store by using set-oriented storage.

I. The mode is free.

J. And supporting dynamic Key-Value inquiry.

K. Full indexing is supported, containing internal objects.

L. support replication and failover.

Use efficient binary data storage, including large objects (such as video, etc.).

And automatically processing fragments to support the expansibility of the cloud computing hierarchy.

And O. supporting JAVA, C + +, PHP, PYTHON and other languages.

P. support online backup, cut and restore by table (set) and library instances.

Q. with visual client management tool.

The data storage system applies the following techniques:

1)Hadoop

hadoop is a distributed system infrastructure developed by the Apache Foundation. The distributed file system is characterized by high fault tolerance, is designed to be deployed on low-cost hardware, provides high throughput to access data of an application program, and is suitable for the application program with a super large data set.

2) Hbase database

Hbase is a high-reliability, high-performance, nematic-oriented and scalable distributed storage system, and a large-scale structured storage cluster can be built on a cheap server by using a related technology.

3)Kafka

Kafka is a high-throughput distributed publish-subscribe messaging system that can handle the flow of actions for large-scale data production and consumption. The activities of data distribution, web browsing and searching are a key factor for many application functions in industrial production and social networking.

4)MongoDB

MongoDB is a database based on distributed file storage. The data structure supported by the method is very loose, and more complex data types can be stored. The most important characteristics are that the supported query language is very powerful and supports the establishment of indexes for data.

(V) a general analysis and calculation module: and (3) establishing an algorithm model base, and realizing real-time, quasi-real-time and off-line calculation by relying on the distributed calculation capability of Storm, Spark and MapReduce.

The cloud computing system is mainly used for carrying out unified data analysis and calculation based on the data computing requirements of a power plant level. The calculation analysis of the system comprises three types of real-time calculation, historical statistical calculation and state calculation.

51) And (3) calculating a real-time index: the method mainly calculates real-time indexes such as alarm point statistics, real-time integral electric quantity and the like. Distributed computing based on Storm.

52) And (3) calculating a statistical index: the method mainly calculates statistical indexes, mainly aims at historical data statistical calculation, and supports various calculation models.

53) And (3) state calculation: the system mainly aims at the fault information of the unit, gives an alarm and displays in real time, and can store information such as data generation, data termination and data grade.

The system realizes the functions of power plant level index analysis, equipment diagnosis, fault early warning and the like by establishing a power plant level basic algorithm library, an expert model library and an intelligent model library and providing means of excavation, development, encapsulation, verification, optimization and the like of a high-level algorithm package. The analysis and calculation system of the power plant comprises the following aspects:

1) and a visual configuration interface is supported, including the establishment of a model library and an algorithm library.

2) And the method supports graphical result display and displays information such as unit performance, fault state, early warning and the like.

3) The method supports the establishment of an online intelligent model, and the model algorithm comprises a neural network, an SVM and the like.

4) And optimization of the online model is supported.

5) The calculation result is stored in a local database.

6) And the online trigger calculation function is supported, and online recalculation, compensation and the like can be realized according to the service requirement.

The cloud system applies to the following technologies:

1)Storm

storm is a distributed, highly fault-tolerant real-time computing system. Can be used to process the continuous messages and save the processed results into the medium. The Storm processing components are distributed, and the processing delay is extremely low, so that the Storm processing system is used as a general distributed framework.

2)Spark

Spark is a big data distributed programming framework, and a system for realizing fast and common cluster computing is realized. The distributed data is abstracted into an elastic distributed data set, so that application task scheduling, RPC, serialization and compression are realized, and an interface is provided for upper-layer components running on the application task scheduling, RPC, serialization and compression.

(VI) an intelligent calculation and analysis module: based on a machine learning algorithm, an intelligent algorithm model is established and trained by utilizing distributed computing resources, and a prediction model is established through data after multidimensional associated data analysis. The intelligent calculating and analyzing module establishes a unit performance model, a thermodynamic system model and an equipment state model according to real-time and historical data, establishes an early warning model according to a monitoring range and an object, associates measuring points, performs rule definition and model training, automatically obtains a residual error initial value and other important default setting parameters through data training after the model is established, can conveniently change the model parameters at any time, records the trace of parameter modification in the system, obtains a variable which cannot be directly obtained through calculating the measuring points, expected values, residual errors and residual error alarms in the original measuring points, the calculated measuring points and the predicted measuring points can be used as input values of the calculated measuring points, the calculated measuring points are positioned under a project and can be quoted by the model and the rules under the project, the configuration function of the calculated measuring points is arranged under the project management, and is configured with the model, The rule configurations are located at the same level. The method specifically comprises the following functions:

data cleaning: to prevent signal failure or incorrect information from being used incorrectly, the acquired signal is first data validated and a validation function is supported that accommodates different periods of time.

Selecting a model: the modeling mode should apply the principles of accuracy and real-time responsiveness, depending on the equipment and system. The system should be flexible, open, configurable. After training, not only all relevant departments can use the model (operation, maintenance, diagnosis experts and the like), but also the model can be maintained independently and a new model can be developed.

Model creation and training: establishing a unit performance model, a main thermodynamic system model and an equipment state model by a second party according to real-time and historical data; according to the monitoring range and the object, an early warning model is established, measuring points are associated, the work such as rule definition and model training is carried out, and the following functions are realized: key equipment and system monitoring models are automatically created from available survey points using a variety of modeling techniques. The normal state of the process and components of the critical equipment and systems of the power plant can be monitored.

Setting model parameters: after the model is created, the system can automatically obtain residual initial values and other important default setting parameters through data training so as to allow the model to be quickly put into use. The user can change these model parameters at any time and conveniently, recording traces of parameter modifications in the system.

Model adjustment and optimization: the equipment and systems of the power plant are subject to constant changes due to technical modifications, unit maintenance and normal aging, and the model should be able to change at any time.

And (3) calculating a measuring point: some variables which cannot be directly obtained, such as time average values, can be obtained by calculating the measuring points. Direct signals, residuals, calculated variables, verified points, etc. may all be used to calculate the signal. Expected values, residual errors and residual error alarms in the original measuring point (origin sensor), the calculated measuring point (ComputedSensor) and the predicted measuring point (StateEstimation) can be used as input values of the calculated measuring point. The calculation measuring points are positioned under the engineering and can be referred by the model and the rule under the engineering, so that the configuration function of the calculation measuring points is arranged under the engineering management and is positioned at the same level with the configuration of the model and the configuration of the rule.

And (3) early warning rule management: and providing management and early warning judgment of rules based on expressions. The rules can be freely defined by signals, residuals and thresholds, and a rule alarm is generated when the behavior of the power plant meets the definition. The expected values, residual errors, residual alarms in the original measuring point (OriginalSensor), the calculated measuring point (ComputedSensor) and the predicted measuring point (StateEstimation) can all be used as input values for the Rules (Rules).

Managing the version of the model: and the edition, operation and isolation of the historical model version are realized through the model version management technology.

(seventh) an edge calculation module: the prediction model becomes an important calculation service node of a big data system by using the calculation capability of the intelligent equipment, and then the intelligent terminal can perform real-time analysis and diagnosis according to the obtained model by issuing the algorithm model through the system. The edge calculation module specifically includes:

71) the measuring data display and analysis module is used for comparing data of the same measuring point at different moments, calculating absolute difference and relative difference of numerical values at different moments, marking differences of different degrees by using colors, displaying time sequence data of the measuring points by using a trend graph, wherein the abscissa is time, the ordinate is the numerical value of the measuring point, the data of a plurality of measuring points are displayed in the trend graph, different measuring points are distinguished by using colors, a continuous curve in a period of time is drawn for a certain measuring point, and the change of a time interval is supported.

72) And the performance online analysis module is used for prompting the running state of the unit and prompting the running mode of the unit at a certain moment. The method mainly comprises the following steps:

and the numerical value comparison table is used for comparing and analyzing different data types of the same measuring point, the data types comprise a measured value, a verification value and an expected value, the comparison result comprises an absolute deviation and a relative deviation, and a large deviation can be marked by colors. The system distinguishes different data types with different suffixes, M represents a measured value, which is a value obtained directly from a sensor or a value calculated from a sensor value; v represents a verification value, which is a calculated value obtained by measuring redundancy by using error transfer and thermodynamic relations according to a measured value and uncertainty; e represents the expected value, which is calculated using the thermal equilibrium under different environmental conditions, assuming the plant is brand new or optimal.

And (4) verifying the measuring points, namely comparing the measured values with the verified values, calculating a punishment value (Penalty) of a certain measuring point at a certain moment, and analyzing according to the punishment value to determine whether the measuring point numerical value is reliable or not, wherein the larger the punishment value is, the more unreliable the measuring point numerical value is.

And loss analysis, wherein the influence of each parameter deviation on the economy in the operation of the power plant is analyzed and displayed, wherein the economical influence comprises output deviation and efficiency deviation.

73) And the performance offline analysis module supports a user to change input data, such as ambient temperature, atmospheric pressure, ambient humidity, condenser pressure, circulating water inlet temperature, fuel low heat value and auxiliary machine power consumption, and calculates various indexes under different input data, such as circulating efficiency, circulating net efficiency, circulating total output and net output, circulating total heat consumption and net heat consumption, by using a thermodynamic model.

74) The operation and maintenance decision support module can provide an auxiliary decision for the operation of the power plant by using the simulator on the basis of index calculation and operation deviation analysis, and can generate an economic report in a customized manner through historical data simulation. The method combines the actual operation conditions (including fuel conditions, actual operation loads, actual unit operation modes and the like) of the power plant, utilizes an optimization algorithm, takes the optimal overall economy of the unit as an objective function, takes the safety of the unit as a constraint condition, carries out automatic optimization calculation on controllable parameters, and provides an optimized operation mode and an optimized effect. Specifically, the air compressor washing prediction is provided, and the system can predict the optimal air compressor washing time according to the running state of the combustion engine and give a corresponding washing effect.

(eight) micro-service release module: the system is used for issuing and managing data bearing, computing tasks and external services of the system. The micro-service publishing module adopts a micro-service technical architecture, each service has a processing and lightweight communication mechanism, and the micro-service publishing module can be deployed on a single or a plurality of servers. The method is mainly characterized by modularization, loose coupling, autonomy, decentralization and the like. The system provides services to the outside through the API gateway. The architecture of the microservice publishing module is schematically illustrated in fig. 5.

The API gateway is used as the only entrance of all clients, receives the client request and routes to the proper service, and opens different APIs according to different clients. The API gateway can realize access control load balance and also can realize security verification that whether the client has the authority to access the service, and the like. Different business services are mutually called through REST, and the services are deployed in a container. The system provides registration and discovery of services, heartbeat monitoring, current limiting, degradation, fusing, and the like. The cache server provides cache service and can also avoid the traffic storm from rushing down the relational database, the business system can store a large amount of data which are accessed or changed quickly in the cache server so as to improve the system access data, and the data in the cache can be flushed into the relational database according to a subscription strategy. The message queue can reduce the coupling between services and can also be used as a buffer between calls, thereby ensuring that the message backlog can not wash down the called party.

(nine) a data governance module: the method comprises the steps of managing main data and metadata, uniformly managing various heterogeneous data, establishing a management system and providing support for application of an upper-layer service system. The architecture diagram of the data governance module is shown in fig. 6. The data management system is a systematic project, and can help an electric power enterprise to standardize a data flow, manage main data in the enterprise, improve the data quality of the enterprise, and ensure that the enterprise obtains accurate, timely and complete data support in business operation management. The main work comprises a data management and guarantee system, a data sharing service system, data full life cycle management, data standard management, data quality management, main data management, metadata management, data security management and the like.

The main data is high-value basic data shared among various systems and repeatedly applied to a plurality of business processes by the power plant, and is a basis for information interaction among various business functions of the digital power plant, such as organization, material coding and the like. The main data management adopts a series of rules, applications and technologies to coordinate and manage basic data related to core business entities of the power plant, provides high-quality main data which is applied across businesses, is consistent in use and is shared for the digital power plant, and reduces cost and complexity.

Metadata is mainly used for describing concepts, relationships and rules related to technology, business and management in the system, and comprises information such as definition of objects and data structures in the data system, mapping of source data to destination data, description of data conversion, indexes, personnel roles, station responsibilities, management processes and the like. Metadata management is planning, implementation and control actions performed for obtaining high-quality and integrated metadata, so that the description and classification of the power plant data information are realized in a uniform format, and the real meaning of the data is clear.

Data governance is a series of specialized work that develops data as an organizational asset, and is the full lifecycle management of the data. The data management system is a system for comprehensively combing, constructing and continuously improving various aspects of data models, data architectures, data quality, data safety, data life cycles and the like of an organization from multiple dimensions such as organization architecture, management system, operation specification, IT application technology, performance assessment support and the like. The data management aims to improve the quality (accuracy and integrity) of data, ensure the safety (confidentiality, integrity and availability) of the data and realize the sharing of data resources in various organizational departments; the integration, butt joint and sharing of information resources are promoted, so that the informatization level of a company is improved, and the informatization function is fully exerted.

This power plant data intelligent processing system builds based on open source Hadoop ecological component, and the use includes: components such as Hadoop, Yarn, Zookeeper, Spark, Storm, HBase, Hive, Kafka and the like can realize real-time data acquisition, historical (discrete) data extraction, data governance, data preprocessing, data storage (structured and unstructured), data analysis and calculation and service release by adopting an SOA service framework design based on the distributed components, break through a one-to-one traditional mode, and efficiently utilize a physical server through a resource allocation and scheduling mechanism at the bottom layer, so that the requirement of running multiple sets (multiple types) of operating systems on one server is met.

This power plant data intelligence processing system carries out the degree of depth through with technologies such as big data, thing networking, cloud computing, artificial intelligence and fuses, can realize the accurate collection of each system data of power plant, realizes the intelligent processing of each system data, improves power generating equipment automation, digitization, visual and intelligent degree and equipment reliability and the availability ratio.

This power plant data intelligence processing system still possesses following advantage:

(1) the method provides a mass, reliable and extensible data storage service, gathers the storage capacity of each node in the cluster, can automatically shield software and hardware faults, and provides uninterrupted data access service for users.

(2) And the incremental capacity expansion and the automatic balance of data are supported, and the random read-write and additional write operations are supported.

(3) By means of a multi-tenant mechanism, each department can independently manage own data. Unless explicitly authorized, the tenants cannot access the other party's data.

(4) The system has a perfect authority authentication and isolation mechanism, provides a multi-level security sandbox, supports read-write authentication, fully ensures the privacy of user data, and avoids data leakage.

(5) A unified authority management model is provided, namely, data authorization inside the tenants or data authorization between the tenants are all in accordance with the same set of authority management mechanism.

(6) Supporting the control of data authority in a database; the transparent encryption of key data is supported, upper-layer application is not required to be modified, and meanwhile, the performance cannot be influenced in the encryption and decryption process.

(7) All main functional modules support multiple user authentication modes such as LDAP and AD, and support SSO authentication.

(8) The isolation management and control of the underlying CPU, memory and disk resources are supported by multiple tenants, and the dynamic allocation of the resources under the condition of the multiple tenants is supported.

(9) The method supports the user to analyze and process the data in the range by using different model algorithms as required, and simultaneously supports the data sharing in a fine-grained authorization mode.

(10) User space protection and cross-user space access authorization are supported, and a plurality of users can collaboratively complete data analysis.

(11) An administrator-oriented unified view of data resources is provided.

(12) Scheduling services are provided for tasks in the cluster system while supporting online services that emphasize response speed and offline tasks that emphasize processing data throughput. The system can carry out uniform scheduling management and operation monitoring on all job tasks in a uniform data system, covers the whole data link (data synchronization, data cleaning, data processing, data analysis and the like), and relates to all distributed computing capacities in the system.

(13) Each component provides a standardized interface and can be well integrated with a third-party product.

(14) The unified data system supports customized development and secondary development so as to quickly meet business requirements.

The above description is only for the preferred embodiment of the present invention, but the present invention should not be limited to the embodiment and the disclosure of the drawings, and therefore, all equivalent or modifications that do not depart from the spirit of the present invention are intended to fall within the scope of the present invention.

Claims

1. An intelligent processing system for power plant data, comprising:

the operation and maintenance management module: the system is used for realizing system physical node management, distributed component management, multi-tenant and user management, data storage and job management;

a safety management module: the system is used for data management of physical security, component security, network security and application software security, forms a multi-dimensional and three-dimensional security support, and guarantees the stable operation of the system and the data security;

a data acquisition and extraction module: the acquisition of the heterogeneous data in the target area is realized through protocol acquisition software based on IEC60870 and an ETL extraction tool;

a data storage module: establishing an operation data storage center, a time sequence data storage center, a data warehouse and unstructured data storage based on an Hadoop ecological open source component;

general analysis and calculation module: and (3) establishing an algorithm model base, and realizing real-time, quasi-real-time and off-line calculation by relying on the distributed calculation capability of Storm, Spark and MapReduce. The method specifically comprises the following steps: the system comprises a real-time calculation module, a historical statistics calculation module and a state calculation module, wherein the real-time calculation module is used for calculating real-time indexes including alarm point statistics and real-time integral electric quantity based on Storm distribution; the historical statistic calculation module is used for calculating indexes for historical data statistics; the state calculation module is used for calculating the fault information of the unit, giving an alarm and displaying in real time and storing the information of data generation, end and grade;

intelligent calculation and analysis module: based on a machine learning algorithm, an intelligent algorithm model is established and trained by utilizing distributed computing resources, a prediction model is established through data after multidimensional associated data analysis, the intelligent computing and analyzing module establishes a unit performance model, a thermodynamic system model and an equipment state model according to real-time and historical data, an early warning model and associated measuring points are established according to a monitoring range and an object, rule definition and model training are carried out, after the model is established, a residual error initial value and other important default setting parameters are automatically obtained through data training, the model parameters can be conveniently changed at any time, the trace of parameter modification is recorded in the system, variables which cannot be directly obtained are obtained through the measuring points, expected values, residual errors and residual error alarms in the original measuring points, the measuring points and the prediction measuring points can be used as input values of the measuring points, the calculation measuring points are positioned under the engineering and can be quoted by the model and the rule under the engineering, and the configuration function of the calculation measuring points is arranged under the engineering management and is positioned at the same level with the configuration of the model and the configuration of the rule.

An edge calculation module: the prediction model becomes an important calculation service node of a big data system by using the calculation capacity of the intelligent equipment, and then the intelligent terminal can perform real-time analysis and diagnosis according to the obtained model by issuing an algorithm model through the system;

the microservice issuing module: the system is used for issuing and managing system bearing data, computing tasks and external services;

the data management module: the method comprises the steps of managing main data, managing metadata, carrying out unified management on various heterogeneous data, establishing a management system, and providing support for the application of an upper-layer service system, wherein the main data management is to adopt a series of rules, applications and technologies to coordinate and manage basic data related to a core service entity of a power plant, provide high-quality main data which is applied across services, is consistent in use and is shared for the power plant, and reduce the cost and the complexity; metadata management is used to describe concepts, relationships and rules related to technology, business and management in a system, including definitions of objects and data structures within the system, mapping of source data to destination data, description of data transformation, metrics, personnel roles, job responsibilities and management processes.

2. The power plant data intelligent processing system according to claim 1, wherein the operation and maintenance management module is a unified visual management and control module constructed based on open source component development, and is open to developers, operation and maintenance personnel and managers, so as to implement unified visual management and control of data, applications and resources, and specifically includes:

infrastructure and component management module: the system is used for installing distributed clusters of Hadoop, Yann, Zookeeper, Spark, HDFS, HBase, Hive, Storm and MapReduce components, realizes the installation and management of the components and the components depending on the components, supports the automatic installation and deployment of the components, uses tools for automatic installation, monitors the operation condition of a big data cluster, including the state and the operation load of a host node, a network node and a frame, simultaneously monitors the occupation condition of software and hardware of each node of the system, and if an abnormal condition occurs to a monitored object, the monitoring system can send an alarm notice on a related management alarm page;

a multi-tenant management module: the method is used for sharing the same system or program in a multi-user environment, and ensuring the isolation of data among users, so that a big data system is fully utilized, service is provided for more users, the data safety of the users and the normal operation of application programs are ensured, and the traceability is realized while resource sharing is realized;

resource and job management module: the method has the advantages that the storage resources are distributed, a visual storage resource distribution interface and a visual storage resource retrieval interface are provided, the computing resources are distributed by the resource scheduling component according to the computing condition, a visual job submission function and a plurality of job arrangement functions are provided, the jobs are conveniently issued and the process is conveniently arranged, meanwhile, a job monitoring interface is also provided, the operation running state and the resource using condition are conveniently known, and the job running log is conveniently checked;

a log management module: combining Logstash, elastic search and Kibana tools to construct a centralized log management system, collecting logs of different components, storing the logs in a centralized manner, and providing a full-text retrieval function;

an alarm management module: the system has a configurable alarm management function and can configure alarms of different levels according to alarm rules; an alarm notification mechanism and an alarm manager can be configured according to actual service requirements.

3. The power plant data intelligent processing system according to claim 1 or 2, wherein the security management module comprises a system security management module and a physical security management module, wherein the system security management module is managed from a host layer, a component layer, a network layer and an interface API layer, the security authentication of the host layer is realized by identity authentication, access control, security audit, intrusion prevention, malicious code prevention and resource control, the security authentication of the component layer is realized by a Ranger framework, the security authentication of the network layer is realized by a Kerberos protocol, and the security authentication of the interface API layer is realized by user login and JWT authentication; the physical security management module manages from physical location selection, physical access control, theft and damage prevention, lightning and fire prevention, water and moisture prevention, static electricity prevention, temperature and humidity control, power supply, and electromagnetic protection.

4. The system for intelligently processing power plant data according to claim 1 or 3, wherein the data collection and extraction module is used for collecting structured, semi-structured and unstructured data generated in the production operation and operation management process of the power plant, collecting real-time production data from each control system according to IEC61850 and IEC61970 standards, extracting discrete historical data from subsystems through ETL extraction tools, and dividing the data according to data types, wherein the collected data mainly comprises production time sequence data, document type data and discrete relational data extracted from other system relational data of the power plant.

5. A power plant data intelligent processing system as claimed in claim 1, wherein the data storage module comprises: the system comprises a time sequence data storage module, a structured data storage module and an unstructured data storage module. The time sequence data storage module is used for storing real-time data which are generated in the production process and stored on the basis of time sequences; the structured data storage module is used for storing operation records and management logs generated in the production or management process of the power plant subsystem; the unstructured data storage module is used for storing information which is irregular and cannot be expressed by a two-dimensional logic table, such as audio, images and HTML, and the unstructured data is stored by adopting a NoSQL database.

6. The power plant data intelligent processing system of claim 1, wherein the edge calculation module comprises: the system comprises a measurement data display and analysis module, a performance online analysis module, a performance offline analysis module and an operation and maintenance decision support module, wherein the measurement data display and analysis module is used for comparing data of the same measuring point at different moments, calculating absolute difference and relative difference of numerical values at different moments, marking differences of different degrees by using colors, displaying time sequence data of the measuring points by using a trend graph, the abscissa of the trend graph is time, the ordinate of the trend graph is measuring point numerical value, a plurality of measuring point data are displayed in the trend graph, different measuring points are distinguished by using colors, a continuous curve in a period of time is drawn for a certain measuring point, and the change of a time interval is supported. The performance online analysis module utilizes the comparison table to compare and analyze different data types of the same measuring point, the data types comprise a measured value, a verification value and an expected value, the comparison result comprises an absolute deviation and a relative deviation, and a large deviation can be identified by colors; the performance offline analysis module is used for supporting a user to change input data; the operation and maintenance decision support module is used for providing an auxiliary decision for the operation of the power plant by using the simulator on the basis of index calculation and operation deviation analysis, and can generate an economic report in a customized manner through historical data simulation.

7. The intelligent power plant data processing system of claim 1, wherein the microservice publishing module employs a microservice technology architecture, each service has its own processing and lightweight communication mechanism, and can be deployed on a single or multiple servers, and an API gateway is used as a unique entry for all clients, receives client requests, and routes them to the appropriate service, the API gateway opens different APIs according to different clients, and implements access control load balancing through the API gateway, and also implements security verification whether a client has authority to access the service, different business services are mutually called through REST, and the services are deployed in a container, and the system provides registration and discovery, heartbeat monitoring, current limiting, degradation, and fusing of the services.