Study On Identification Algorithms Of Systems With Abrupt Parameter Changes

With the development of economy, science and technology, the level of automation has been greatly increased. This requires the corresponding improvement of system identification technology, a discipline on modeling. It's still a difficult problem to track the time-varying system because of the complexity of time-varying system and the limitation of people accessing infonnation from it. In this dissertation, some problems of system identification have been studied. The researches are focused on tracking problems of time-varying systems. Several fast tracking algorithms have been presented. The dissertation is composed as follows:1.A new algorithm, incremental estimation RLS algorithm with variable parameters and the local polynomial approximation is presented.The concept of error levels is given , the method to obtain the error levels and .the ranges of parameters ale also given.This algorithm has fast tracking ability and good ability to overcome the noise.7A new algorithm, gradient estimation algorithm with variable parameters and the local polynomial approximation is presented.This algorithm is a quasi-robust algorithm.The method to obtain the error levels and the ranges of parameters are given. This algorithm has fast tracking ability and good ability to overcome the noise.3.A new algorithm, recursive robust minmax estimation algorithm with moving time window is given.The principle to obtain the valve value is also given.How to obtain the algorithm parameters is also given. This algorithm has fast trackingHIability.4.System identification algorithm based on the improved GAs is presented.The criteria to obtain the algorithm parameter is given. It remains the main adavantages of GAs and has fast tracking ability and good ability to overcome noises.5.A combined algorithm of improved GAs and RLS algorithm with dead zone is presented. A combined algorithm of recursive robust minmax estimation algorithm and RLS algorithm with dead zone is also presented.These two combined algoritluns both remain the adavantages of the RLS algorithm with dead zone and improve their tracking ability.