This paper presents a computationally efficient Multi-Objective Dynamic Programming (MODP) algorithm. The algorithm is applied to obtain the optimal supervisory control for PHEVs to minimize two objectives -total CO2 emissions and operational dollar costs to an individual PHEV owner. The algorithm integrates the concept of crowding distance from the Multi-Objective Evolutionary Algorithms (MOEA) literature. This distance metric is used to refine the optimal Pareto front at every time step for each state discretization. The refinement of the Pareto front significantly reduces the computational time and memory required for MODP, making it feasible. At the same time, the results show that the refinement retains optimality and produces a Pareto front with a good spread ranging from one extremal point to the other. The results also reveal interesting insights for the tradeoffs that can be achieved in minimizing the CO2 emissions and cost objectives for the underlying grid mix and driving conditions assumed.