智能电网监控中的分布式复杂事件处理技术研究

发布时间：2018-05-02 04:27

  本文选题：智能电网 + 事件驱动计算　； 参考：《华北电力大学(北京)》2017年博士论文

【摘要】：智能电网将逐步推广分布式能源高渗透率、电动汽车与双向互动技术,涌现的大量主动实体及其智能监控系统将涉及海量测量流式数据、实时数据交换通信与系统高度复杂性,建立实时在线监测分析、高反应性与可扩展的大数据流式计算管理平台尤为关键。本文分析了智能电网分布式实时监控特征,数据密集型监测大数据在线实时分析应用需求,应用以数据为中心基于发布/订阅(P/S)的实时消息中间件,建立高度分布、异构、动态与自组织的主动实体监测数据流在线实时收集与共享的分布式大数据管理应用架构;基于复杂事件处理(CEP)研究智能电网高性能流式计算、模式匹配与主动响应决策应用技术。主要工作如下:(1)提出建立专业化的智能电网监控分布式实时大数据管理系统,以满足能源物联网中物理上高度分布、异构、动态与自组织的自治实体为核心的实时在线监测计算与反应性控制需求。基于最新通用大数据流式计算架构模式分析,提出面向智能电网监控系统专业化的事件驱动的分布式流式数据处理应用技术架构,采用基于复杂事件处理技术(CEP)的多源事件流实时在线查询和模式匹配的处理层与基于P/S的DDS通信服务的消息层的分层独立设计模式,构成分布式CEP的流式计算体系,同时给出其关键实现技术与需求侧双向互动应用的案例分析。(2)在ICT中,智能电网在线监控应用属于分布式实时系统集成,其实时通信技术成为研究焦点;本文提出了基于DDS建立智能电网实时P/S通信服务的架构模型与信息安全部署设计。基于DDS的智能电网实时通信服务需要遵循IEC 61850标准体系,针对智能电网IEC 61850通信标准的抽象通信服务接口(ACSI)具体映射实现的不一致和复杂性的难点问题,本文应用DDS实时中间件支持通用化的基于P/S模式的IEC61850通信实现方法,提出模型驱动软件开发(MDSD)的基于DDS的IEC 61850ACSI的统一映射实现方法,支持Qo S通信策略的动态配置;并给出智能变电站GOOSE多播通信案例分析与实验验证。面向IEC 61850元模型的DDS映射实现具有简捷实用性,由自描述SCL配置语言易于实现通用化的ACSI通信统一动态配置管理。进一步工作将开发基于DDS的用户配置实用化软件。(3)本文梳理了分布式CEP事件模型与应用体系;提出适合智能电网监控的语义CEP事件本体模型与语义查询处理流程;以及主动配电网(ADN)分层监控中基于DDS的分布式多CEP Agent服务的应用部署方法。事件模型标准化是支持对事件的采集、通信、存储与分析,提高电力系统运行动态反应性的基础性问题。电力系统动态交互中存在海量数据与事件的信息交换,目前规范了CIM数据交换标准,但表达动态特性的事件交换标准仍没有制定。本文研究建立智能电网语义CEP系统的标准化的事件本体模型;作为案例分析,设计了AMI监测类事件本体知识库系统,实现对监测事件流的主题、时间与空间的语义及其关联的知识表示,提高语义CEP基于知识库的模式匹配与查询能力。CEP是流式计算的高级模式,支持多源事件流模式匹配与主动响应,符合智能电网监控系统对运行状态与行为在线分析的特定需求。本文梳理了CEP核心事件模型、模式匹配查询语言与规则,以及分布式CEP服务架构模型,明确了架构中通信接口对系统可扩展性的支持。提出了以数据为中心的DDS通信服务标准化接口支持分布式CEP应用部署;针对分布式能源高渗透率的ADN分层分布式监控模式,部署可随需定制的层次化CEP Agent节点实现主动实时监控管理,DDS数据总线支持CEP Agent部署的异步性、并发性与自组织特性,以建立灵活的可扩展的高反应性的主动实时监控应用,并给出了面向全局数据空间(GDS)的CEP多类型事件主题注册管理与案例分析。(4)研究了智能电网用户需求侧双向互动关键应用技术问题,本文提出了自动需求响应(ADR)系统中基于DDS/CEP的分布式监控服务集成架构,以及与用户双向互动的需求响应性能实时在线监测分析方法;并开发了分布式CEP用户响应性能计算仿真实验进行验证。ADR是体现智能电网与用户双向互动,适应分布式能源高渗透率的最佳应用。本文基于Open ADR通用事件模型和IEC PC118智能电网用户接口标准,开展了ADR系统的双向互动技术的应用研究。以用户基线负荷与需求响应性能实时在线监测分析为案例,定义CEP服务节点功能;设计ADR事件驱动的需求侧事件模型及查询算法;提出扩展CEP引擎的实时与历史数据流混合查询算法及其性能实验测试,开发了分布式的ADR CEP用户响应性能在线计算仿真实验。进一步工作将探索在大规模快速ADR中分布式CEP监控服务集成的实际应用潜力。(5)提出数据驱动的智能电表非线性负荷序列分段预测新颖算法。目前基于核方法(KM)的非参数模型是继人工神经网络方法的富有前景的短期负荷预测方法。本文提出模糊C聚类与基于KM的非参数模型的组合预测算法,并采用某地区实际负荷数据的详细实验与误差分析,验证组合模型具有较高的预测精度。模型简单可靠、训练参数少、具有良好的时间性能、自适应性强,可推广应用到智能电网监测时序大数据短期预测分析。
[Abstract]:Smart grid will gradually extend the high permeability of distributed energy, electric vehicle and two-way interactive technology, a large number of active entities and their intelligent monitoring systems will involve mass measurement of flow data, real-time data exchange communication and high complexity of the system, real-time online monitoring and analysis, high responsiveness and extensible large data flow meter. The computing management platform is particularly critical. This paper analyzes the characteristics of distributed real-time monitoring of smart grid, the application requirements of data intensive monitoring data online real-time analysis, and the application of data centric P/S based real-time message middleware to establish a highly distributed, heterogeneous, dynamic and self-organizing active entity monitoring data stream online. Real-time collection and sharing of distributed large data management application architecture; based on complex event processing (CEP) research on smart grid high performance flow computing, pattern matching and active response decision application technology. The main work is as follows: (1) a specialized distributed real-time large data management system for smart grid monitoring is proposed to meet the energy resources Federation. The real-time on-line monitoring calculation and response control demand at the core of the physical highly distributed, heterogeneous, dynamic and self-organizing autonomous entities. Based on the latest general large data flow computing architecture model, an event driven distributed flow data processing application technology architecture oriented to the specialization of the smart grid monitoring system is proposed. The hierarchical independent design pattern of multi source event flow real-time online query and pattern matching based on complex event processing technology (CEP) and the message layer of P/S based DDS communication service is designed to form a flow computing system of distributed CEP. At the same time, the case analysis of the key implementation technology and the two-way interactive application of the demand side is given. (2) ICT The online monitoring and application of smart grid belongs to the distributed real-time system integration, and the real time communication technology is the focus of research. This paper proposes a framework model and information security deployment design based on DDS for the establishment of real-time P/S communication service in smart grid. The DDS based real-time communication service of smart grid needs to follow the IEC 61850 standard system and aim at the intelligence. The abstract communication service interface (ACSI) of the power grid IEC 61850 Communication Standard mapped the difficult and difficult problem of the inconsistency and complexity. This paper applies the DDS real-time middleware to support the universal IEC61850 communication implementation method based on the P/S mode, and proposes a unified mapping implementation of the IEC 61850ACSI based on DDS in the model driven software development (MDSD) development (MDSD). Methods, support the dynamic configuration of Qo S communication strategy, and give the case analysis and experimental verification of GOOSE multicast communication in intelligent substation. The DDS mapping for the IEC 61850 element model is simple and practical, and the unified dynamic configuration management of ACSI communication is easy to be realized by the self description SCL configuration language. Further work will be developed based on DDS The user configuration practical software. (3) this paper combed the distributed CEP event model and application system, proposed semantic CEP event ontology model and semantic query processing process suitable for smart grid monitoring, and the application deployment method of distributed multi CEP Agent service based on DDS in active distribution network (ADN) stratified monitoring. Event model standard It is the basic problem that supports the acquisition, communication, storage and analysis of events, and the basic problem of improving the dynamic response of the power system. There is a exchange of information between massive data and events in the dynamic interaction of the power system. At present, the standard of CIM data exchange is standardized, but the event exchange standard that expresses the dynamic characteristics is still not formulated. This paper studies the establishment of intelligence. The standard event ontology model of the power grid semantic CEP system is standardized. As a case analysis, the AMI monitoring class event ontology knowledge base system is designed to realize the theme of the monitoring event flow, the semantic and related knowledge of the time and space, and to improve the semantic CEP based on the pattern matching and query ability of the knowledge base, which is the high flow calculation. The level mode supports multi source event flow pattern matching and active response, which conforms to the specific requirements for the online analysis of the running state and behavior of the smart grid monitoring system. This paper combs the CEP core event model, the pattern matching query language and rules, and the distributed CEP service architecture model, and makes clear that the communication interface is extensible to the system in the architecture. A data centric DDS communication service standardization interface is proposed to support distributed CEP application deployment. A hierarchical and distributed ADN distributed monitoring model for distributed energy high permeability is proposed, and a hierarchical CEP Agent node is deployed to implement active real-time monitoring and management, and DDS data bus supports the asynchronism of CEP Agent deployment. Concurrency and self-organizing features, in order to establish flexible and scalable and highly reactive active real-time monitoring applications, and give a CEP multi type event topic registration management and case analysis oriented to global data space (GDS). (4) the key application technology problems of the two way mutual interaction in the demand side of the smart grid users are studied. The automatic requirement is proposed in this paper. The distributed monitoring service integration architecture based on DDS/CEP in response (ADR) system and the real-time online monitoring and analysis method of demand response performance in two-way interaction with users; and developing a distributed CEP user response performance simulation experiment to verify that.ADR reflects the two-way interaction between the smart grid and the users and adapts to the high penetration of distributed energy. Based on the Open ADR universal event model and the IEC PC118 smart grid user interface standard, this paper has carried out the application research of the two-way interactive technology of the ADR system. Using the real-time online monitoring and analysis of the user baseline load and the demand response performance as a case, the definition of the node function of the CEP service and the design of the demand side event driven by the ADR event is designed. Model and query algorithm; propose an extended CEP engine real-time and historical data flow hybrid query algorithm and its performance test test. A distributed ADR CEP user response performance online calculation simulation experiment is developed. Further work will explore the practical application potential of distributed CEP monitoring service integration in large-scale and fast ADR. (5) data The non parametric model based on the kernel method (KM) is a promising short-term load forecasting method based on artificial neural network. This paper proposes a combination pre calculation method of fuzzy C clustering and KM based non parametric model, and uses the actual load data in a certain area. The detailed experiment and error analysis show that the combined model has high prediction accuracy. The model is simple and reliable, and the training parameters are few. It has good time performance and strong adaptability. It can be applied to the short term prediction analysis of the time series of smart grid monitoring.

【学位授予单位】：华北电力大学(北京)
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TM76
，

本文编号：1832361