A robust mean variance optimization model is set up for an economic dispatch problem integrating uncertain wind power in a real-time electricity market. The objective of the model is to find a robust optimal solution so as to minimize both the expected total generation cost and hedge the risk resulting from the robustness. Furthermore, the proposed model is transformed into a biobjective second-order cone programming. To generate an evenly distributed Pareto front set that will not overrepresent one region of the design space or neglect others, we propose a normalized-constraint method in this work. The Pareto front, demonstrated in the Institute of Electrical and Electronics Engineers 118-bus case study, shows a trade-off between total generation cost and risk. An important observation during the case study is that the Pareto front of the robust mean variance model has a special shape that, by sacrificing only a little total generation cost, can greatly reduce the risk in the robust real-time economic dispatch.