Local energy communities (LECs) are viewed as key actors in the future clean energy system. However, uncertainty is inherent in renewable energy generation and electricity consumption. Renewable energy sources (RES) like solar and wind can exhibit considerable variation in power generation depending on weather and other factors. Optimisation techniques like Robust Optimisation (RO) aim to take uncertainty into account for decision making. It is important to consider uncertainty in both domains - power supply and electricity demand to facilitate the participation of LECs in the clean energy transition. In this paper, we examine empirical time-series data from the building energy management system (BEMS) of a public university in Ireland (University College Dublin) to characterise the temporal correlation between photovoltaic (PV) power generation and electricity consumption. We propose a data-driven approach to construct correlated uncertainty sets for supply and demand data by partitioning the day into specific time-windows. This data-driven approach can reduce the level of conservatism in RO models for LEC distribution network design and operational decision problems.