## Introduction

Throughout the world, electrical utilities include a hefty charge in a facility's bill for the peak electrical demand incurred during the billing period, usually a month. Such a charge is part of the utility's cost recovery or amortization of newly installed capacity and for operating less efficient power plant capacity during higher load periods.

Demand charge is a good portion of a facility's electrical bill. Typically a demand charge can be as much as 50% of the bill, or more. Thus, to reduce the demand cost, many industrial and commercial facilities try to "manage their loads." One example is by moving some of the electricity-intense operations to "off-peak" hours"â€”when a facility's electrical load is much smaller and the rates (\$/kW) are lower. But, when moving electrical loads to "off-peak" hours is not practical or significant, a facility will likely consider a set of engine-driven or fuel cell generators to run in parallel with the utility grid to supply part or all the electrical load demand during "on-peak" hours. We call these Peak Shaving Generators or PSGs.

While the electric load measurement is instantaneous, the billing demand is typically a 15-to-30- minute average of the instantaneous electrical power demand (kW). To obtain the monthly demand charge, utilities multiply the billing demand by a demand rate. Some utilities charge a flat rate (\$/kW-peak per month) for all months of the year. Other utilities have seasonal charges (i.e. different rates for different seasons of the year). Still, others use ratchet clauses to account for the highest "on-peak" season demand of the year.

Thus, the model presented in this paper focuses on the development of a method to obtain the optimal PSG size (g*kW) and PSG operation time (hours per year) for a given facility. This model is for the case of a facility with a constant billing demand rate (\$/kW/month) throughout the year. The analysis is based on a linear load-duration curve and uses a simplified life-cycle-cost approach. An example illustrates the underlying approach and optimization method. In addition, the paper shows an EXCEL spreadsheet to implement the optimization model. We call this model PSG-1.

0 0