Modeling and Analysis of Load Ensembles

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

Power systems strive to maintain balance between electricity production and consumption so that the system frequency remains close to its nominal value. This has traditionally been achieved by varying generation to track changes in the total load. Increasing reliance on renewable generation is, however, challenging this operating strategy. Renewable generation is inherently variable and offers only limited ability to assist in meeting the supply-demand balance. Conceptually, this reduction in generation control can be compensated by introducing fast-acting control of loads. The project will consider modeling and control strategies for harnessing the potential of large numbers of small loads (load ensembles) to assist in regulating the power system frequency. Of particular interest are strategies that are non-disruptive, in the sense that individual consumers are oblivious to changes in their loads, yet the aggregate response plays a vital role in grid operation. Such aggregate load control is appealing in that it facilitates significant expansion of renewable generation by utilizing existing resources (loads) rather than requiring extensive (and expensive) new infrastructure such as bulk energy storage. The research for controlling aggregate loads would impact grid operations, increasing system reliability and energy efficiency. The proposed work will include integrating research results into courses and training of graduate students as well as dissemination at workshops and conferences. In addition, technology transfer to a load aggregator startup company will be explored. The project will explore two forms of aggregate models for load ensembles. The first captures the aggregate flexibility across an ensemble of loads, quantifying the controllability that is available from those loads. This can be thought of as building an equivalent energy-storage (battery) type representation. The work will rely on zonotope and homothet concepts to establish polytopic approximations of ensemble availability and flexibility. The so-called bin model will also be considered as a means of assessing ensemble dynamics and designing control schemes. Investigations will consider the influence of noise and heterogeneity on the aggregate dynamics of load ensembles. The project will develop systematic approaches for incorporating those effects into the bin model, and will explore the error introduced by the approximations inherent in the bin representation. Situations have been observed where control of load ensembles results in undesirable, highly nonlinear behavior such as sustained oscillations and bifurcations. Such effects arise due to unanticipated synchronization of large numbers of participating loads. The project will use of a hybrid (switched) dynamical systems framework to explore these nonlinearities and develop strategies for circumventing such behavior. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.