Collaborative Research: EAGER-DynamicData: Machine Intelligence for Dynamic Data-Driven Morphing of Nodal Demand in Smart Energy Systems

Purdue University, West Lafayette IN

Investigators

Abstract

The electric power grid is the indispensable infrastructure for power delivery and distribution. It is a system of high complexity and heterogeneity comprised of a variety of interconnected systems, subsystems, generators, and loads. In addition, it is a dynamic system with evolving characteristics that suffers from several infrastructure limitations, which if not handled properly, may lead to instabilities with severe consequences including costly brownouts and blackouts. However, advancements in information and data-driven technologies offer the necessary ground for developing tools that efficiently monitor grid infrastructure and manage electricity flows in ways that achieve and maintain high performance and reliability in grid operations. Towards that end, coupling power systems with information systems converts traditional electric energy delivery infrastructures into interconnected hybrid energy-data systems called smart energy systems, where power flow is controlled via information signals. Dynamic data available in smart energy systems includes, but is not limited to, hourly user energy consumption measurements from smart meters, electricity pricing signals, system voltage readings from GPS-synchronized measuring units scattered throughout the network that can take hundreds of readings per second, and data from weather stations. Thus, due to grid complexity, a tremendous amount of information is not only generated but also transferred throughout the grid, and grid participants, such as customers, utility companies, and grid operators, are exposed to multiple heterogeneous data streams coming from various sources. In this data intensive environment, participants are being engaged to make fast real-time decisions regarding morphing of their load demand and consumption behavior patterns. Nodal load forecasting is identified as a key point for developing future smart energy systems and electricity markets. The principal theme of this research is the fast and optimal nodal load morphing in smart energy systems that takes into account big volumes of dynamically varying data. In particular, this research addresses the problem of management and processing of big data within the framework of Dynamic Data Driven Systems (DDDS) as applied to nodal load morphing. The focus of this study will be the development of a set of new intelligent and self-adaptive algorithms for online big data processing and fast real-time decision-making in smart energy infrastructures. The main feature of the current research is the integration of machine learning DDDS with dynamic optimization methods to solve the computational problem of forecasting optimal or near-optimal shapes of a load in a timely manner accounting for multiple streams of continuously incoming data and their inherent uncertainty. Emphasis will be given in handling and processing incentive signals and more particularly electricity pricing signals as a major factor in load morphing. Furthermore, extensive testing and verification of the developed algorithms will be performed on real-time simulated scenarios obtained with the GridLAB-D software simulator. In short, the proposed research for nodal load morphing will enable a new and transformative approach towards efficient, inexpensive, and fast processing of big data as applied to smart energy systems.