Abstract: To obtain the mesoscopic model of composites, an approach on parameter identification of composite components with high accuracy was proposed. Based on the finite element model of micro unidirectional carbon reinforced polymer composite (CFRP), the sensitivity matrix of static displacements with respect to elastic parameters of composite components was formulated and the objective function was defined as the 2-norm of differences between the measured and calculated data on displacements. In order to overcome the problem caused by the magnitude differences between identified variables, the relative sensitivity was chosen to improve the precision and efficiency of parameter identification. The fiber uniform distributed composite plane model and fiber random distributed 3D model were employed respectively to verify the validity and accuracy of the components parameter identification methods. In addition, the influences of number of measuring points and measurement errors on the parameter identification method were revealed. Results show that the identification method of components parameters of composite materials in the presented study is stable considering the influences of number of measuring points and measurement errors.