For cyber-physical systems(CPS) with complex cause-effect logic relations between multiple subsystems, a CPS multi-cause system model with unknown nonlinear and uncertain coupling is established. A distributed control-cause method is also proposed based on cloud control technologies. A nonlinear decoupled observer is designed to achieve dynamic feedforward linearization for the CPS multi-cause system which is decomposed into multiple CPS cause systems without coupling. By the nonlinear decoupled observer, both a distributed model predictive controller and a distributed optimization algorithm are given to realize online constrained optimization control for the multiple CPS cause systems. Effectiveness of the proposed control strategy and optimization algorithm is verified by a numerical simulation.