Classification of Infrequent Disturbances

Abstract

Infrequent disturbances will not necessarily be picked up in an identified noise model. Yet their presence may be important for controller design. This paper suggests a technique to classify different types of infrequent load disturbances.It is assumed that a model for the plant is available or can be identified from data. Given this model, the predicted output can be compared with data to infer the disturbance. In this paper a bank of Kalman filters is used to estimate the disturbance sequence. The suggested technique has been successfully tested via simulations and laboratory experiments.