fax 01-972-952-9435. AbstractReservoir engineering requires to manage many sources of uncertainties that can be classified in two categories : (a) uncontrolled uncertainties on the physical reservoir description parameters and (b) controlled uncertainties on the reservoir development scheme parameters. In this context engineers must answer several problems : describe prior uncertainties, identify the ones that actually influence the oil production process, make safe production forecasts and optimize the reservoir production scheme.In this paper, several statistical methods dealing with these problems are presented. They are (a) Experimental Design, (b) Response Surface Methodology and (c) Monte-Carlo methods. Integrating these techniques enables to build a simplified model of a process and to estimate the uncertainties on the response predictions. The entire procedure was applied to a field case showing both types of controlled and uncontrolled uncertainties. The result is a new frame allowing engineers to quantify uncertainties on the reservoir production forecasts conditionally to uncertainties on the reservoir modeling parameters. The procedure has lower computational costs than the traditional one but is quite complex and needs adapted software to be used by reservoir engineers.

TX 75083-3836, U.S.A., fax 01-972-952-9435. AbstractWhile designing Deep Offshore fields developments, several decisions have to be made : defining the system architecture, selecting technologies, defining operating and maintenance policies, etc. The objective is to ensure production by mitigating risks while lowering costs and maximising profits.However, risks are numerous (e.g. equipment failures or Flow Assurance issues) and strongly interact with each other thus making the behaviour of the production process quite complex. Hence, in order to make correct decisions, engineers need a tool able to :− integrate various risks, especially those arising from Flow Assurance issues, − simulate accurately the complex behaviour of the subsea production system all along the field lifecycle, − estimate the performances of the candidate designs in order to choose the best one. This paper describes a risk management methodology based on the general framework provided by Dependability. After a hazard identification step, a model is built based on hybrid interpreted stochastic Petri nets which allow to model complex interactive systems and mix both discrete (e.g. equipment failures) and continuous (e.g. progressive degradations such as corrosion or deposits) aspects of the production process. The model provides fast computations of the availability and production availability of the system thus giving criteria for risk management.The methodology was applied to a simplified representative subsea production system and allowed to quantify the influence of major risks in terms of economic consequences and optimize a maintenance policy.

No abstract