The Art and Science of Energy Forecasting

Abstract

Distinguished Author Series articles are general, descriptive representations that summarize the state of the art in an area of technology by describing recent developments for readers who are not specialists in the topics discussed. Written by individuals recognized as experts in the area, these articles provide key references to more definitive work and present specific details only to illustrate the technology. Purpose: to informthe general readership of recent advances in various areas of petroleum engineering. Introduction In the 1970's, energy forecasting moved into the limelight. Market forces, which had evolved during extended industry growth, combined with politicalevents to trigger two disruptions to the normal flows of petroleum. These "oil shocks" raised the awareness of politicians and the general public to thesignificant role of energy supply in the modem economy. Suddenly, energy-particularly oil-was everyone's business. Almost everyonehad a complaint and a prescription. "Experts" appeared as if by magic, each prescription. "Experts" appeared as if by magic, each suggesting a different path out of the wilderness. At the core of the uncertainty were some basic economic and forecasting problems. Are we really running out of oil? Do oil and energy obey the laws ofsupply and demand? Is energy demand price elastic? Can a cartel really work forlong? Is it possible to understand all of the ramifications of a particular policy? Can history be used to help predict the future? policy? Can history beused to help predict the future? As the answers evolve, we begin to understand how much still remains to be learned about energy forecasting. There is a large body of literature about forecasting theory and forecasting techniques, of which the References provide only a small sample. But ratherthan survey the repertoire of econometric techniques including linear programming, multiple regression, Box-Jenkins techniques, input-output models, cross impact analysis, and the Delphi approach, this paper focuses on philosophical problems posed by any attempt to forecast the future. At the outset, note the distinction between two classes of future occurrences: events and trends. Events, which cannot be known before they happen, include things like storms, earthquakes, and assassinations. Even thespecific outcome of a process (for example, which person will be electedpresident in 1988) is an event that cannot be foretold. On the other hand, trends lend themselves to analysis and, thus, can beforecast within the constraints of assumptions about those events and trendswhich are external to the analysis. Examples include sales of products, long-term weather patterns, and political and social trends. Although it is impossible to predict who will be elected president in 1988, it is possible to analyze and forecast the trends which may lead to the predominance of aparticular party or philosophy. party or philosophy. Trend forecasting must also be distinguished from path forecasting. Here the distinction is usuallyone of path forecasting. Here the distinction is usually one of the timehorizon involved. The inertia of any system tends to keep it in motion along aparticular trend even though the specific path may be quite irregular and volatile. The trend can be forecasted; the path is unpredictable. Most forecasts are of trends and attempt to define the broad movements thatwill later be called "history." An example of a trend forecast is shown in Fig.1. (All figures provided by Chevron Corporation Economics Department.) The trend assumes that real oil prices will increase at a defined rate until they are high enough to stimulate substantial quantities of synthetics. On any given date, however, the market price of oil is likely to deviate because of prevailing price of oil is likely to deviate because of prevailing marketconditions. Although the path will generally follow the trend, it will besubject to temporary deviations and excursions. The challenge is to determine whether a particular price move indicates a trend or whether it is a temporary excursion outside the normal range of forecasting uncertainty. Beware the Ides of March -William Shakespeare History of Forecasting Forecasting-as opposed to fortune-telling-was largely a creation of theIndustrial Revolution. JPT P. 1437