Research On Grid Dispatching Decision Support System

Energy Management System plays a more and more important role in modern grid control center. However practical requests demand EMS performing more analysis under various state beyond normal state. Technology of communication and computer enable more plentiful and accurate data to be sent to control center via a faster and more flexible means. Thus, more computer decision support results can be achieved at the control center based on the newly delivered data. In this environment, this paper presents a notion of constructing a decision support system for grid dispatch center, plus fault handling and some dynamic analysis function to EMS. More AI techniques are used in the dispatch decision support system to provide better decision results to the dispatchers. Considering security and stability of grid as the primary goal, the paper presents some methods on building the system, especially on fault handling and transient stability assessment.Ontology is used to ensure open characteristics to the decision support system. Through building field ontology for power system dispatch analysis, software modules can share more knowledge and implement more effective communication beyond pure data. Thus, software has more clear structure and behaves better characteristics of openness.To reuse existing EMS resource and be able to run automatically, the system is constructed with a Multi-Agent type architecture. The paper describes the structure and running logic of single agent, communication means among agents, structure and function of environment that agents live in, as well as the perceiving means of agents in environment. Through modeling of environment and single agent, the ability of automatically run is achieved.As to fault handling, a reasoning framework for power system restoration is built with Petri-Net technique and is used in searching for restoration plan. In this framework, all structures and behaviors of elements in the restoration process are described with unified language, and the framework has some advantages on modeling concurrency, resource limiting and collision, as well as advantages of faster searching speed than Expert System. Another advantage of this framework is it decreases the complexity of model by object-oriented technique when handling large-scale system.In dynamic analysis, a new data mining method combining machine learning and information fusion is proposed to use the plentiful data acquired from on-line record and off-line simulation about transient stability. The new method can be used a fast algorithm for...