The need to advance the state of the art in smart grid technologies is increasingly recognized by scientists, engineers, practitioners and policy makers, as the critical mechanism to improve the energy efficiency of producing and using electricity in our homes, businesses, and public institutions. The smart grid enhances the traditional electricity distribution system with three critical components: (1) two-way communications of energy and digital information (thus enabling the combination of energy and information as entities being exchanged), (2) monitoring infrastructures to supply information to the communication network, and (3) computational intelligence to use such information to maintain stable optimal operational efficiency and energy delivery and to enable autonomous response to abnormal or disruptive events. The smart grid allows the control and monitoring of the entire grid, from the utility to the smart home—including intelligent appliances and plug-in electric vehicles. The overall goals are energy conservation, cost reduction, and enhanced reliability, security and transparency.
The energy delivery paradigm, as explored through the projects of iCREDITS, is driven by an anticipation of better quality and larger quantities of data communicated to agents/entities spread over distribution and transmission networks. This monitoring project adds real time situational awareness, that must be communicated to the agents. This allows the agents to modify their operation, control and wide-area protection or emergency response strategies, in order to accommodate the new constraints introduced by a disturbance. This typically results in a sudden change in topology, and may sometimes threaten the stability of system. In normal operating conditions, the aim of the agents would be to re-optimize their objectives, to incorporate the change; in stressed operating conditions, the aim would be to secure the power system from wide scale outages.
This project will establish a solid foundation for real-time disturbance data analysis in power systems. It will identify existing, and develop new data mining and real time classification theories. These theories will be particularly suited for the substantially high volume disturbance data, including bad data and many unidentified events, sampled at high sampling rates by synchronized measurement devices in power transmission and distribution systems. In particular, we will develop a framework with the following components: (1) Feature extraction techniques through extensive comparison of existing approaches, (2) Semi-supervised learning clustering approaches to learn unknown disturbance events, (3) unsupervised discriminative pattern discovery algorithms to characterize different disturbances, and (4) real-time classification and localization of disturbances.