Real-time simulation training system for substation based on field-programmable gate array (FPGA) is proposed in order to reduce the construction cost of substation simulation training system and enhance the quality of real-time digital simulation. A minimum degree maximum independent set method is adopted to arrange the sequence of nodes elimination and voltage computation, which has a good balance of computation burden and degree of parallelism. Fine granularity parallel computing is realized by adopting a multi-input and multi-output instruction stream arithmetic unit, therefore improving FPGA resource utilization. The simulation parameters are indirectly modified by the status word and effect word, thus reducing the simulation time and saving data memory space. Simulation example shows that a simulation substation with 542 nodes could be processed in an EP4CGX150 chip with 40 ms simulation step, according to the proposed minimum degree maximum independent set method, instruction stream arithmetic unit and indirect modification method for simulation parameters. This work is supported by National Natural Science Foundation of China (No. 51477114) and Tianjin Science and Technology Program (No. 13TXSYJC40400).