The lack of monitoring data in Low Voltage (LV) networks has lately become a major concern since their operation is currently undergoing significant changes driven by the worldwide desire to support and facilitate the energy transition. It is therefore essential to improve network observability, which leads to the deployment of smart metering (SM) devices at the end-user level. However, their actual roll-out is confronted to technical, financial as well as social barriers, and is therefore still limited to some sparse areas. This paper aims at overcoming this data deficiency in the context of long-term studies of the system. The objective is thus to establish reliable individual stochastic models for every LV consumers (e.g. residential, commercial, etc.) and distributed generators, even those without metering devices. The first step of the work focuses on the segmentation of end-users into representative clusters. Afterwards, within each cluster, all available SM information is used to extrapolate statistical profiles of all components thanks to an innovative load modelling methodology. The accuracy of our implemented approach is then validated on a real LV feeder thanks to a Monte Carlo simulation. In particular, the improvement of this new modelling method compared to currently used approaches, such as Synthetic Load Profiles, is statistically highlighted.