Parameter estimation in switched systems

Parameter estimation in switched systems is an interesting topic, since such vital models can be used for modeling a wide range of real life systems controlled by a hysteresis or other switching mechanics. In this work, a model for enabling the extended Kalman filter to estimate the parameters in such a switched hysteresis system is presented and tested. The nonlinear model used in the filter is found by a continuous approximation of the discrete switching function used for describing the switched systems. The proposed filter are tested which shows that the filter is able to estimate all the model parameters, this includes the parameters describing the switching. And since the EKF is a recursive filter the method is suitable for real life implementation, therefore a test on an actual data set from an operational turbine have been used with success to validate that the proposed method is actually applicable for parameter estimation in a real world switched system.