Date of Graduation

12-2014

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Engineering (PhD)

Degree Level

Graduate

Department

Electrical Engineering

Advisor/Mentor

Juan C. Balda

Committee Member

Simon Ang

Second Committee Member

Christophe Bobda

Third Committee Member

H Alan Mantooth

Fourth Committee Member

Roy A. McCann

Keywords

Distribution Systems, Energy Storage, Markov Models, Photovoltaic Generation, Power System Optimization, Scaling Methods

Abstract

Energy storage units (ESU) are increasingly used in electrical distribution systems because they can perform many functions compared with traditional equipment. These include peak shaving, voltage regulation, frequency regulation, provision of spinning reserve, and aiding integration of renewable generation by mitigating the effects of intermittency.

As is the case with other equipment on electric distribution systems, it is necessary to follow appropriate methodologies in order to ensure that ESU are installed in a cost-effective manner and their benefits are realized. However, the necessary methodologies for integration of ESU have not kept pace with developments in both ESU and distribution systems. This work develops methodologies to integrate ESU into distribution systems by selecting the necessary storage technologies, energy capacities, power ratings, converter topologies, control strategies, and design lifetimes of ESU. In doing so, the impact of new technologies and issues such as volt-VAR optimization (VVO), intermittency of photovoltaic (PV) inverters, and the "smart" PV inverter proposed by EPRI are considered.

The salient contributions of this dissertation follow. A unified methodology is developed for storage technology selection, storage capacity selection, and scheduling of an ESU used for energy arbitrage. The methodology is applied to make technology recommendations and to reveal that there exists a cost-optimal design lifetime for such an ESU. A methodology is developed for capacity selection of an ESU providing both energy arbitrage and ancillary services under a stochastic pricing structure. The ESU designed is evaluated using ridge regression for price forecasting; Ridge regression applied to overcome numerical stability and overfitting issues associated with the large number of highly correlated predictors. Heuristics are developed to speed convergence of simulated annealing for placement of distributed ESU. Scaling and clustering methods are also applied to reduce computation time for placement of ESU (or any other shunt-connected device) on a distribution system. A probabilistic model for cloud-induced photovoltaic (PV) intermittency of a single PV installation is developed and applied to the design of ESU.

Download

Share

COinS