Using an Optimization Software to Lower Overall Electric Production Costs for Southern Company

Abstract

The electric utility business has traditionally been very capital-intensive. In the last 10 years, however, fuel costs have escalated at a faster rate than all other cost components. To minimize its fuel cost expenditures, Southern Company installed a comprehensive operational-planning software package to forecast system loads, optimally schedule thermal and hydro units, and estimate future prices of power transactions. The heart of the package is the Wescouger optimization program, designed by the Advanced System Technology division of ABB Power Systems, Inc. By using state-of-the-art dynamic programming and branch-and-bound techniques, Wescouger has helped Southern Company save over $140 million in fuel costs over the past seven years. Today, this optimization software is one of the key scheduling tools used at the control center in Birmingham to fulfill the pool's most important responsibility—to deliver electricity to customers in the most reliable and least costly manner.

Strategic Operations Research and the Edelman Prize Finalist Applications 1989–1998

“…1990, Brigandi et al 1994, American (Anbil et al 1991, Smith at al. 1992, Vasquez-Marquez 1991 and Delta Air Lines (Subramanian et al 1994), Harris Corporation (Leachman et al 1996), National Car Rental (Geraghty and Johnson 1997), Sadia (Taube-Netto 1996), Sears (Weigel and Cao 1999), and Southern Company (Erwin et al 1991) all used OR to tackle very complex operations problems over a period of time, with each new development adding to the sustainability of the advantage created. ABB Electric (Gensch et al 1990) and Vilpac Truck Company (Nuno et al 1993) used OR to assist or lead major organizational change that would be costly for competitors to replicate quickly.…”

Power Systems Restructuring

“…A simplified version of the general unit commitment problem formulation results when the optimization is performed assuming that the demand is equal to the forecasted value (47,50]. In this case a random load Pd,sys(t) is replaced by its predicted value Pd,sys(t).…”

“…Many unit commitment methods in use are deterministic with respect to the load, meaning that the optimization is performed assuming that the actual demand Pd,sys is equal to its forecasted value Pd,sys (47,50]. It can be shown that a schedule obtained by replacing a randomly varying demand by its expected value would be a suboptimal method, sometimes being very far from optimal, when applied to the stochastic unit commitment problem formulation given in equation (2.40) above.…”

“…Note that an optimal policy need not be unique. 50. For this formulation, we assume that the generator is capable of selling as much power as desired at the market equilibrium price Pk· For analyzing the impact of generation constraints on decision making and a numerical example illustrating intertemporal effects , see [36].…”

“…Reference [50] presents a model for energy brokerage/auction system with reserve margin and transmission losses. This chapter uses the described framework in reference [50] and the proposed model in reference [51] to develop an auction market simulator.…”