This paper describes an information system (STEPS) designed to support the identification of ill-defined systems, and subsequent use for prediction of their behaviour. Ill-definedness is brought about by unavoidable inadequacies in model structure, usually in conjuction with sparse and unreliable empirical data. The uncertainty modelling used in STEPS is based on set-theoretic concepts, i.e. the uncertainties are expressed in terms of bounds, and not in terms of statistical parameters. The set-theoretic framework is outlined briefly. To assist the identification STEPS also contains recursive parameter estimation tools based on the stochastic concept rather than the set-theoretic concept. STEPS also provides support tools for data management, for model structure improvement and for the construction of predictions with the model. The information systems is demonstrated by applying it to the identification of a simple dissolved oxygen model for a lake.
- System Identification
- Uncertainty analysis