Electricity capacity planning requires simultaneous consideration of multi-decadal and sub-annual power sector dynamics along with interactions of the power sector with other economic sectors and uncertain physical conditions. A challenge for long-term multi-sector models is to capture critical structural elements of the power sector without drastically increasing solution complexity. We incorporate sub-annual dynamics in the power sector module of a multi-sector model (GCAM-USA). We combine linearly optimal dispatch modeling for electricity generation with a non-linear method for new capacity investment assuming no foresight. With this approach, we are able to model the impact of future surprises. We demonstrate this capability by modeling the impact of different natural gas price paths on state-level power-sector investment and operation in the U.S.