Search

With the rapid proliferation of cloud computing, the energy cost of datacenters and the impact of reactive power on the grid have become major concerns. This paper presents two energy management control schemes for distributed uninterruptible power supply (UPS) architectures utilizing lithium-ion ultracapacitors (LIC) for improved system efficiency and reactive power mitigation. One scheme is proposed for low-power servers with an internal dc bus, while another is proposed for medium-to-high-power server racks with an in-rack dc bus. The LIC and the dc bus interface with a bidirectional dc–dc converter, which allows the power supply unit to run in the optimal operating region for the improved system efficiency and power factor. The server-level UPS architecture is experimentally tested using a 200-kHz, bidirectional, multiphase dc–dc converter with hysteretic current-mode control, while achieving 33% reactive power mitigation. Based on the SciNet datacenter in Toronto, ON, Canada, a rack-level distributed UPS architecture is built in MATLAB, and system-level simulations indicate a 37% reduction in the reactive power. [ABSTRACT FROM AUTHOR]

The peak current estimation of lightning detection networks for strikes to tall towers is discussed in this paper. Such systems are sometimes calibrated using return-stroke current data obtained by means of rocket-triggered lightning or instrumented towers of relatively short height. However, for strikes to electrically tall towers, they tend to overestimate the return-stroke current peak. In this case, in fact, the associated radiated electromagnetic fields, from which the return-stroke current is estimated, experience a significant enhancement with respect to the field that would be radiated if the same return stroke was initiated at ground level or on a short tower. Two approaches to correct the current estimates of a lightning detection network for a lightning strike to a tall tower are discussed and applied to the current measurements obtained at the CN Tower in Toronto in the summer of 2005, for which estimates were available from the North American Lightning Detection Network (NALDN). It is shown that correcting the NALDN estimates using the so-called tower factor obtained from theoretical studies results in an excellent estimation of lightning current peaks. [ABSTRACT FROM AUTHOR]