Study Of System Identification For Missing Data And Related Software Design And Implementation

System identification theory has been considerable development, and the various branches are alsomaturing. But the general identification methodsare considered in the condition that the data is full, identification document involving missing data is not much.This paper briefly introducesthe system identificationtheory, and different identificationalgorithmsare designed in diffrenentcases. In the case of frequency-domain with full data, bode diagram and Matlab toolbox are designed to identify themodel order andparameters of the system at the same time. In the case of the unchanged time-domain with full data, offline and online identification algorithmsare used to getthe system’s order andparameters simultaneously. In the case ofthe unchanged time-domain with missing data,the auxiliary model is designedfirstly, andthe current model parameters are identified based on previous input and output information. In the case ofthe unchanged time-domain, there are two identification methods:1)recursive least squaresalgorithm:theassistance model is designed firstly, and then the variable forgotting factor is introduced which is changing according to the situation of missing data and variable parameters; 2)improved Kalman filter identification algorithm:assistance model is designed firstly, and then parameters noise matrix is setted according to the order of the system, so Kalman filter identification algorithm can be tracked in real time to the changing parameters; 3)identification algorithm on adjustable parameters’order:assistance model is designed firstly, then adjust the parameters’ order based on the data error, so the algorithm can improve the accuracy of the results.Finally, the paper describesthe design of the software system. According to the different types of data, the software system selects the corresponding identification function interface. With Visual Studio 2010, Matlab 2012b and SQL SEVER 2008, the whole software system is integrated by these functions. So, the speed of the system is improved, and the users can see identification parameter results directly which is drawn by Matlab.