Low-voltage distribution network (LVDN) often faces information omissions issues like unknown accurate topology and impedance parameters, hindering effective voltage control of LVDN with rapid integration of household photovoltaics (PVs). This paper proposes a data-driven topology and parameter estimation method using the measurement data from smart meters. The phase of each user is identified based on sensitivity features, and three sensitivity criteria are proposed for LVDN topology identification with large number of latent nodes. Based on the bottom-up sweep-backward method, impedance parameters are estimated, and the network loss compensation is considered to further improve the identification accuracy. Then, a voltage control model of LVDN is established to improve the user voltage quality under household PV access. Finally, the effectiveness and advantages of the proposed method are validated by numerous simulations and comparisons in a practical LVDN in China, the proposed method shows better identification performance in LVDN with large-scale latent nodes, and demonstrates good performance in voltage control.