FORECASTING PETROLEUM PRODUCTION USING CHAOS TIME SERIES ANALYSIS AND FUZZY CLUSTERING

Abstract

Forecasting of petroleum production time series is a key task underlying scheduling of oil refinery production. In turn, forecasting requires analysing whether time series exhibits chaotic behavior. In this paper we consider chaos analysis based forecasting of time series of gasoline and diesel production. Chaos analysis is based on Lyapunov exponents and includes determination of optimal values of embedding dimension and time lag by using differential entropy approach. For forecasting of petroleum production, fuzzy IF-THEN rules constructed on the base of fuzzy clustering of the time series are used. The obtained prediction results show adequacy of the used methodology.