Millimetre wave (mmWave) signal is promising for the challenge of bandwidth shortage and can motivate the research on large-scale antenna arrays. In this paper, the authors investigate the hybrid precoding design that combines with a radio frequency (RF) precoder and a digital precoder for multi-user multiple-input single-output (MU-MISO) systems with large-scale antenna arrays. In practical large-scale antenna array systems, the small antenna spacing and the properties of a scattering environment can create a correlation between channel coefficients for separated receive units. Such correlation prejudices the performance of multi-user systems. This study proposes two approaches to reduce the channel correlation by the hybrid precoding structure. The first approach selects some vectors with the smallest influence with each other from all array response vectors to design an angular-based RF precoder. The second approach optimises the singular values characteristic of the virtual channel matrix before digital processing for the RF precoder. Then, the MU-MISO baseband precoding is implemented by reduced RF chains. Numerical results show that they have different performance depending on the environment. The first can achieve better performance when the channel correlation is not too high, while the second is more effective when the channel realisation is terribly ill-conditioned, especially for higher signal-noise ratio (SNR).