This paper addresses the operational planning of an ethylene production plant. The problem is formulated as a non-linear programming (NLP) model. The NLP is validated on a real-world petrochemical plant whose main product is ethylene as well as several other hydrocarbon products that are obtained using naphtha as feedstock. The amount produced of each product can be adjusted, within certain bounds, through the operation of specific units, the production of gasoline blends or the tuning of cracking severity. This operational flexibility requires the use of optimization techniques for the production planning. The objective is to maximize the difference between product revenues subtracted from feedstock and utilities costs. The resulting model is non-convex and involves bilinear terms for the flow rates of individual components and their ratios. The model is applied and analyzed under typical scenarios found in the operation of the plant, such as processing naphtha with different properties and formulating gasoline blends, among others. Finally, the model is coupled with a short cut model to determine more precisely utility consumption. This is illustrated with the simulation of a xylene tower that calculates steam consumption for this unit. (c) 2007 Elsevier B.V. All rights reserved.