One key design decision for photovoltaic (PV) power plants is to select the number of PV modules connected in series, also called the string size. Longer strings typically lower total system costs, but the string size must still meet relevant electrical standards to ensure that the maximum system voltage remains less than the design voltage. Traditional methods calculate string size using the temperature coefficient of open-circuit voltage assuming that the coldest expected temperature occurs simultaneously with a full-sun irradiance of 1000 W/m2 . Here, we demonstrate that this traditional method is unnecessarily conservative, resulting in a string size ~10% shorter than necessary to maintain system voltage within limits. Instead, engineers may determine string size by modeling open-circuit voltage over time using historical weather data, a method consistent with the 2017 National Electric Code. For bifacial systems, we derive a simple additive term that predicts the additional voltage rise. We demonstrate that this site-specific modeling procedure predicts open-circuit voltages in close agreement with data from field measurements. We further perform a comprehensive sensitivity analysis to identify an appropriate safety factor. By using site-specific modeling instead of traditional methods, we estimate a ~1.2% reduction in levelized cost of electricity, a significant improvement toPV power plant economics. The method is provided as an easy-to-use web tool and as an open-source Python package.