An analysis of short-term risk in power system planning.

Abstract

The results presented in this dissertation are as follows: The impact of weather forecasts on Bayesian load forecasting as a function of forecast lead time is shown in Chapter 2. Risk in the presence of load forecast uncertainty alone and risk in the presence of load forecast uncertainty together with fuel price uncertainty are shown in Chapters 4 and 5, respectively. The expected cost of uncertainty, in these chapters, is given as a function of lead time in $/MWh.

The goal of this research is to formulate and present a methodology that evaluates short-term risk in power system planning. Specifically, this research shows how to determine the risk of short-term planning in the presence of electrical load forecast and fuel price uncertainty, both of which have a large impact on the outcome of power system production cost planning. The uncertainty in the load is described by Bayesian forecasting and fuel price uncertainty is modeled by conditional triangular probability distributions.

Classical decision analysis forms the backbone of the methodology presented herein. Throughout this dissertation, sampling theory, load forecasting theory and general engineering are applied with the aim of transforming the short-term power system planning problem into a suitable structure for decision analysis. Probabilistic sampling is used to discretize the load and fuel prices. Then an electrical power production simulation model results in a unit commitment strategy and a cost of each plan. A best, i.e., minimum cost, plan can be selected and the expected cost of uncertainty can be estimated.