Improvement And Application Of Imperialist Competitive Algorithm

Imperialist competitive algorithm (ICA) is a new evolutionary algorithm, it has beenextensively investigated and widely applied in controller design. Traditional PIDcontroller play an important role in control area, and fractional order PID is an extensionof the PID controller using fractional calculus. The operation process of actual system isoften disturbed, and contains the unmodeled dynamics, uncertainty of parameter andstructure, which greatly reduces the control performance of PID controller or fractionalorder PID controller. Aiming at this problem, the grey prediction theory and ICA are usedto optimize the design of the PID controlle and fractional order PID controller in thispaper. This paper fulfills the following work:First, a grey PID controller based on ICA is proposed. The proposed controller adoptsGM(0, N) grey model to estimate the uncertainties and disturbance in the system, and thena grey compensation controller is constructed based on the estimation results to eliminatethe effects of uncertainties. The grey compensation controller together with aconventionalPID controller are called grey PID controller. In order to get better control performance,grey PID controller’s parameters are optimized through the imperialist competitivealgorithm(ICA). The proposed controller is used to control the automatic voltageregulator(AVR) and has achieved good control effects.Second, since the existing fractional order PID controllers are almost structed withthe current control error, this method may lack of foresight, so as to make the controlperformance is poor. Therefore, this paper presents a grey fractional order PID controllerwhich is based on grey prediction theory. The output of a system is predicted by using agrey prediction model GM(1,1) in this method. Then we use the predicted error instead ofthe error item in the traditional feedback control to structe the fractional order PIDcontroller. Meanwhile, we will apply ICA algorithm to find the appropriate parameters ofthe proposed controller for the desired system. The proposed method generates the controloutput according to the changing trend of the future, and has show good preview controleffect.Fiinally, a novel of improved imperialist competitive(ICA) algorithm is proposed.ICA is a new meta-heuristic optimization algorithm inspired by the socio-politicalbehavior in real world. To improve the global exploration ability of ICA algorithm andspeed up its convergence, two improved schemes for ICA algorithm are proposed. The first scheme presents a new possession probability in the imperialistic competition phase.Inspired by “Geopolitics”, the new possession probability takes not only the power of theempire but also the distance between the imperialist and the colonies into account. Thesecond scheme introduces the wavelet mutation operator into original ICA algorithm toimprove its exploration ability. The improved ICA (MICA) algorithms are tested onseveral benchmark functions and used to tune the parameters of fractional orderPID(FOPID) controller. Results show that the MICA algorithms outperform the originalICA algorithm interms of solution quality and convergence speed.