Evolutionary grey-box modelling for practical systems

A novel grey box modelling methodology combining advantages of both black and clear boxes is proposed. The technique makes the best use of a priori knowledge on the clear box global structure of a physical system, whilst it incorporates accurate black boxes for unmeasurable local nonlinearities. Through hybrid genetic evolution and Boltzmann learning, it enables dominant structural modelling with local parametric tuning, without the need for linear parametrisation. Validation results show that the proposed method offers robust, uncluttered and accurate models for two practical systems. It is expected that this type of grey box model will accommodate many practical systems.