Time series analysis of commodities price

The feasibility study of a chemical plant is based on the assessment of both CAPEX (i.e. capital expenditures) and OPEX (i.e. operating expenditures) terms. As far as OPEX terms are concerned, besides the utilities contribute, it is necessary to determine the price/cost of raw materials and sub/products that belong to the more general domain of commodities. The conventional approach to feasibility study assumes constant the commodities price throughout the life of the chemical plant. Actually, a chemical plant can operate for decades. It is common to have chemical plants well maintained and revamped that can work for 20-30 years. A so long time interval calls for a dynamic approach to the economic assessment of the price of commodities to get a more suitable and consistent feasibility study. The advanced approach to feasibility study is based on the so called Dynamic Conceptual Design, which lays its foundations on the time series analysis of commodities price. This analysis allows identifying appropriate regression models to be used in the following feasibility study to track dynamically the OPEX terms, thus achieving a more consistent and reliable economic assessment. The paper focuses on the detailed analysis of historical trends of commodities price to identify both qualitatively and quantitatively the cyclic, seasonal, and stochastic contributes that play a role in forecasting the quotation trends. A reference component (e.g., crude oil) is isolated to assess if the price of other commodities (directly/indirectly derived from the reference one) can be determined according also to its evolution. The time horizon of the study is rather long so to enclose unexpected bullish and bearish periods that characterized erratically the last decades. The paper shows how the price dynamics of commodities and specifically of crude oil is a Markov process characterized by a stochastic and time discrete contribute. It also proposes some Autoregressive with Distributed Lag, ADL, models that are specially oriented to long-time predictions as opposite to more common (but quite different) models developed to evaluate the commodity shocks on short time horizons (i.e. usually few months for the so called futures assessment)