The coupling pattern of a cyber-coupled power system is defined by the way power nodes and cyber nodes are connected. In this paper, we study the effect of coupling patterns on robustness and functionality of cyber-coupled power systems. In a cyber-coupled power system, the cyber nodes to be coupled with power nodes are regarded as decision-making cyber nodes that are selected using a measurement called average propagation latency. Basically, a coupling pattern is formed by ranking the node criticality of the coupled cyber and power nodes, and then connecting them by a specific sequence. We introduce a parameter, called relative coupling correlation coefficient, to quantify the coupling pattern. A lower relative coupling correlation coefficient generally indicates a lower assortativity of coupling, which leads to a degradation of the functionality of coupled systems. Preliminary results show that a coupled system of a lower relative coupling correlation coefficient has better robustness. The finding indicates that increasing coupling assortativity and improving the robustness of a coupled system, are two conflicting objectives. Thus, a multi-objective problem is formulated and the Pareto-optimal solution is derived to balance the two objectives in the optimization of coupling patterns.