Abstract:

 A subspace identification algorithm in the time-domain based on short memory principle is given for multivariable fractional order systems. This algorithm can effectively identify the coefficient matrixes and the fractional differential order of multivariable fractional order systems. Utilizing the Poisson moment functionals (PMF), input and output signals are filtered, and the basic input and output data matrixes in subspace method constructed. The problem of identifying fractional differential order is transformed into a parameter optimization problem according to the cost function. Using the short memory principle of fractional differentiation, the fractional differential order is separated from input and output data matrixes, and differentiation calculation of input and output signals is avoided. A numerical example is given to validate the proposed algorithm.

Key words: fractional order systems, subspace method, short memory principle, system identification