Abstract: The problem of fractional order system identification in time domain is studied,and a new identification algorithm based on subspace method in time domain is presented.When the fractional differential order is known,the parameters of system are identified by utilizing subspace method after calculating the fractional differential of input and output signals and constructing a new input and output data equation.When the fractional differential order is unknown,the problem of order identification is transformed into parameter optimization problem by utilizing cost function.The Poisson filter is used to efficiently avoid the problem that input and output signals should have high order derivative during the calculation of fractional order differential.And the selecting method of weight matrix,which improves the accuracy of identification,is presented through analysis.The numerical simulation validates the algorithm.