How to choose Pseudopotential for DFT calculation in Quantum Espresso?

I think you should start from a better understanding of the physical approximations behind any pseudopotential based electronic Density Functional Theory (DFT) calculation, whatever package is used for computations.

Pseudopotentials are not there for running scf loops, but to describe the bare electron-ionic-core interaction in a way consistent with the idea that the lowest lying one particle Kohn-Sham orbitals are the same as those in an isolated atom or ion. Pseudopotential theory can be found in good modern solid state physics textbooks or in a few review papers in the literature. Quantum Espresso documentation provides a few pointers to some background material. An excellent starting point I would recommend to study, even if you do not plan to generate a pseudopotential by yourself is this paper by Fuchs and Scheffler. It should complemented by some more up-to-date review, but it contains a very detailed description of the underlying theory.

Here, I can try to provide you the general picture, but for details, I recommend you carefully study the basic theory behind norm-conserving pseudopotentials.

In brief, norm-conserving pseudopotentials allow to eliminate the lowest lying electronic shells in favor of an effective interaction between valence electrons and ionic cores, within a DFT description of the electronic properties. This last specification means that one has to interpret one-electron wavefunctions as ingredients to represent the electronic valence density in the Kohn-Sham theory. The basic request for norm-conserving pseudopotentials is that the pseudopotential is able to reproduce exactly the Kohn-Sham all-electron wavefunctions outside a suitably chosen sphere around the nucleus. It is possible to show that such a choice ensures that the pseudopotential-interacting ion has the same scattering properties as a nucleus dressed by its core electrons in a range of energies around a reference level.

Depending on the precise way the pseudopotential is generated, this range of energies may be wide or narrow. In the former case one says the pseudopotential is more transferrable than in the latter case.

Moreover, since pseudopotentials are generated within DFT, they also differ with respect to the DFT approximation they are based on.

All these ingredients of pasudopotential theory have to be at least approximately understood if one would like to understand the approximations a DFT calculation like those Quantum Espresso or other similar packages are based on. Similarly for choosing among different pseudopotentials.

In some cases it could be possible to skip a reasonable understanding of the theory behind, but only if somebody else is using his/her expertise to advise one pseudopotential or another.

Once the theory behind pseudopotential is clear, it is possible to learn how to generate a new pseudopotential from scratch. In the long run, it would be the best choice.