Desert steppe is one of the most vulnerable and severely disturbed ecosystems by humans in terrestrial ecosystems. The accurately simulating carbon and water flux of the desert steppe and its response to anthropogenic disturbance can not only reveal its complex ecological process but also provide a decision-making basis for the artificially ecological conservation and restoration. Ecological models can effectively simulate the carbon and water cycle processes of terrestrial ecosystems, but the numerous parameters of the model and the rationality of these values limits their universal application. Therefore, exploring parameter optimization is an effective way to promote the universal application of the ecological models. In this paper, the physiological and ecological parameters of the Biome-BGC model are optimized by using the Parameter ESTimation (PEST) method and data observed by Eddy Covariance. On the basis of evaluating the effect of parameter optimization, gross primary productivity (GPP) and evapotranspiration (ET) of the planted shrub encroachment ecosystem of the desert steppe region in the Yanchi County of Ningxia from 1986 to 2018 were simulated. The results show that:(1) the parameter optimization can improve the simulation effect of the Biome-BGC model on GPP and ET of planted shrub encroachment ecosystems in the desert steppe region. The simulated GPP and ET after parameter optimization are closer to the observed values, and the simulation effect at the monthly scale is better. (2) The parameter optimization method of the Biome-BGC model based on PEST has strong universality, and the optimized parameters can be applied to long-term GPP and ET simulations of the planted shrub encroachment ecosystems in the desert steppe region. (3) The GPP of the planted shrub encroachment ecosystem of desert steppe region in the Yanchi County of Ningxia showed a slow upward trend from 1986 to 2018, with an increase of 1.47 g C m^-2 a^-1, and the interannual variation rate of ET was large, and there was no significant change trend.