The operating modes of elevation equilibrator electro-hydraulic system are harsh, and many system parameters are soft variables with time-varying characteristics, such as servo valve gain coefficient, bulk elastic modulus, and comprehensive leakage coefficient. They are sensitive to changes in oil pressure, oil temperature, valve opening and system operating points, and the presence of these soft variables poses great difficulties in establishing accurate nonlinear mathematical models of the controlled system. The system state equation was linearized by multiple linear regression method. Then a particle swarm optimization algorithm which is improved successfully identifies the parameters mentioned above offline. The initial values of the servo valve gain coefficient, bulk modulus of elasticity, and comprehensive leakage parameters obtained from the system are significant for the design and determination of subsequent adaptive controllers.