Aim of this paper is to present an application of the reconciliation and optimisation package ORO (On-line Reconciliation and Optimisation) to an hydrogen plant of an existing refinery. The industrial plant performances optimisation can be divided into three phases : gross errors detection, reconciliation and optimisation. All the difficulties linked to each of these phases and the general approach are presented. The model reconciliation and optimisation skills were tested by comparing the predictions with plant measurements for two different plant capacity conditions. Finally different optimisation strategies are shown, for different production targets.