Recent developments in mathematical techniques have allowed the development of a software package to simulate and optimise the flow of hydrocarbons through oil and gas production and distribution networks of arbitrary connectivity and complexity. It has been successfully applied to a wide variety of operational problems including de-bottlenecking, optimisation of compressor strategies and determining the optimal lift-gas allocation to networks of gas lifted wells. This paper discusses the innovative mathematical concepts at the heart of the algorithm taking account of both optimisation and modelling aspects. It demonstrates how an integrated approach to optimisation and simulation based on Sequential Linear Programming (SLP) techniques provides the user with substantially enhanced flexibility and reliability compared with traditional nodal analysis type approaches and how these benefits can be realised in concrete practical examples. Descriptions of the behaviour of physical objects in a production or distribution network is commonly given in the form of measured data sets or discrete performance tables rather than analytical functions. A final focus of this paper is the discussion of how such data sets can be used directly and efficiently in an optimisation context. Introduction Throughout field life hydrocarbon production facilities need to be designed, built, maintained and altered to meet the developing needs of the hydrocarbon reservoirs to which they are connected. Computer models of the behaviour of the facilities are an essential pre-requisite to the efficient management of this process. In this paper we introduce a simulation and optimisation method for hydrocarbon production networks based on a technique called sequential linear programming (SLP). In the sections below we first introduce the concept of SLP and how it can be applied to hydrocarbon production. We then proceed to discuss the key advantages of our technique over other more traditional methods. Additionally, we point out how recent developments in mathematical modelling and optimisation have allowed us to enhance the robustness and speed of our approach. Since its commercial release in early 1999 our technique has been successfully applied to a wide variety of production system design and optimisation problems. For this paper we have chosen two recent case studies. The first is a de-bottlenecking study to enhance the production rates achieved from a medium sized gas gathering network. The second study demonstrates how the software package can be used to optimise the usage of produced associated gas in a highly complex multiphase hydrocarbon production system. Sequential Linear Programming (SLP) Sequential linear programming is a family of techniques for the solution of large-scale non-linear optimisation and simulation problems. The philosophy of the approach was first introduced by Griffith and Stewart who applied it to hydrocarbon processing systems.1 For a recent description of the approach see e.g. Bazaraa et al.2

TX 75083-3836, U.S.A., fax 01-972-952-9435. AbstractRecent developments in mathematical techniques have allowed the development of a software package to simulate and optimise the flow of hydrocarbons through oil and gas production and distribution networks of arbitrary connectivity and complexity. It has been successfully applied to a wide variety of operational problems including de-bottlenecking, optimisation of compressor strategies and determining the optimal lift-gas allocation to networks of gas lifted wells. This paper discusses the innovative mathematical concepts at the heart of the algorithm taking account of both optimisation and modelling aspects. It demonstrates how an integrated approach to optimisation and simulation based on Sequential Linear Programming (SLP) techniques provides the user with substantially enhanced flexibility and reliability compared with traditional nodal analysis type approaches and how these benefits can be realised in concrete practical examples. Descriptions of the behaviour of physical objects in a production or distribution network is commonly given in the form of measured data sets or discrete performance tables rather than analytical functions. A final focus of this paper is the discussion of how such data sets can be used directly and efficiently in an optimisation context.