The use of simulation models in the design and verification of complex industrial systems is becoming standard practice. Indeed, ever-increasing demands for better performance, smaller size and less cost lead to highly integrated systems of components that often have been designed in isolation for a given set of initial and boundary conditions. When integrated into a federated system the system frequently exhibits complex dynamic behavior that had not been anticipated during the design process and may be difficult to control. All models of physical processes are uncertain either due to unknown/ignored phenomena or inherent uncertainty due to variations in operating conditions and/or physical properties. In this talk we discuss several issues pertaining to model based analysis and design of complex systems: 1) We will briefly outline a research program for coordinated control of multiple CHP systems as distributed generation resources and some technical problems that need to be addressed. 2) We will present a fuel cell power plant as an example of a system level dynamic model that was built using state of the art modeling environment and successfully used for design validation and control system design of a highly integrated fuel cell power plant design. 3) We will discuss a model-based analysis of unexpected behavior in a transcritical heat pump. In particular, an undesirable multiplicity of stable solutions (one very efficient and one inefficient) in the heat pump were observed in prototype units. Control oriented modeling, originating from first principles highlights the state-dependent heat transfer coefficient in the evaporator dynamics as a contributing cause to this bi-stable phenomena. Specifically, the bilinear nature of the controlled gas cooler and its coupling to the dynamic non-linearity in the evaporator induces a system-wide bifurcation in the equilibrium conditions. Model results are presented to illustrate this, along with steady-state and dynamic data to confirm the accuracy of the model. 4) We will present a unified (abstract) framework based on concepts from the theory of random dynamical systems for studying parametric and initial condition uncertainty in complex dynamical systems. The notion of input measure of an observable is defined and its propagation to output measure of the observable is studied by means of transfer operators. Uncertainty of these measures is defined in terms of their cumulative probability distributions. The developed formalism is illustrated through an analysis of the effect of pitchfork bifurcation on uncertainty. General results on uncertainty for dynamical systems on an infinite time horizon are derived.