Parameter estimation and calibration by use of exergy

Abstract

Previously, it has been demonstrated that structural dynamic models, i.e. models that are able to account for the change of properties of the species by adaptation of to the prevailing conditions or for a possible change in species composition can be developed by use of exergy as a goal function (see, for instance, Jorgensen, 1997. Integration of Ecyosystems Theories: A Pattern.Kluwer, Dordecht). The exergy is currently computed for the model. The parameters are currently changed, for instance, up to ±10% per week, if the exergy of the model increases by this change. This approach to structural dynamic models is based on the hypothesis that exergy measuring the distance from thermodynamic equilibrium expresses the ‘fitness height’, which is the basis for the ongoing selection. If the hypothesis is valid, it should, however, be possible also to use the hypothesis to find the parameters that are unknown or cannot be found by a normal calibration due to lack of knowledge or observations. A case study will be presented where exergy has been used for a partial calibration of parameters in a eutrophication model. By use of a combination of the usually applied calibration based on minimizing the difference between model results and observations and a calibration based on exergy optimization, it is shown that an improved parameter estimation can be achieved. It is recommended to consider to apply the method, when our parameter knowledge is limited. This method of exergy optimisation may also be utilized as a genetic algorithm in combination with the artificial neural network approach.