This article reports on efforts to extend a MARKAL energy model for South Africa to include rural energy choices, allowing for computation of optimal energy systems in a typical (non-electrified) rural village. A previous study (Howells et al. 2002) highlighted deficiencies in earlier efforts to build models of rural household energy behaviour, such as inadequate calibration against surveys of actual energy use in rural settings as well as limited time resolution. The present study incorporates a new village energy survey. It also deploys TIMES, an extension of the MARKAL computational framework that allows explicit modeling of time-of-day load curves, for demand side management analysis, and the representation of storage devices and end-use technologies ("appliances") that meet more than one energy service concurrently.