The interest in self-consumption of electricity generated by rooftop photovoltaic systems has grown in recent years, fueled by decreasing levelized costs of electricity and feed-in tariffs as well as increasing end customer electricity prices in the residential sector. This also fostered research on grid-connected PV-battery storage systems, which are a promising technology to increase self-consumption. In this paper a mixed-integer linear optimization model of a PV-battery system that minimizes the total discounted operating and investment costs is developed. The model is employed to study the effect of the temporal resolution of electrical load and PV generation profiles on the rate of self-consumption and the optimal sizing of PV and PV-battery systems. In contrast to previous studies high resolution (10 s) measured input data for both PV generation and electrical load profiles is used for the analysis. The data was obtained by smart meter measurements in 25 different households in Germany. It is shown that the temporal resolution of load profiles is more critical for the accuracy of the determination of self-consumption rates than the resolution of the PV generation. For PV-systems without additional storage accurate results can be obtained by using 15 min solar irradiation data. The required accuracy for the electrical load profiles depends strongly on the load profile characteristics. While good results can be obtained with 60 s for all electrical load profiles, 15 min data can still be sufficient for load profiles that do not exhibit most of their electricity consumption at power levels above 2 kW. For PV-battery systems the influence of the temporal resolution on the rate of self-consumption becomes less distinct. Depending on the load profile, temporal resolutions between 5 min and 60 min yield good results. For optimal sizing of the PV power and the storage capacity a resolution of 60 min is found to be sufficient. For the sizing of the battery inverter power of the storage system, a finer temporal resolution of at least 300 s is necessary. (C) 2016 Elsevier Ltd. All rights reserved.