Long-term, integrated modelling of electricity generation capacity and its operational requirements is important for coherent energy policy because of the multi-decadal lifespan of such plant, and because related fuel requirements, water consumption, and greenhouse gas emissions impact other sectors. This paper describes a physical model of electricity generation designed for long-term simulations, including the reproduction of decades of historical data, and its application to key policy questions. The simulation uses a dynamic, physical input-output model for the basic material and energy industries, embedded in an economy-wide stocks and flows framework (SFF). The historical reproduction, illustrated for the state of Victoria, is based on integrating bottom-up technical data on physical processes as inputs, and validating output against top-down aggregate data. We show that while greenhouse gas emissions would decrease significantly with the aggressive introduction of cleaner generation technology, long-term reductions are difficult to maintain against ongoing growth in consumption. Additionally, our sensitivity analysis identifies the importance of rapid action due to the age structure of operating power stations. We also identify significant imposts of alternative power generation on land and water resources, though a business-as-usual scenario would impose greater losses of water resources.