For simulation of grid-connected PV system with solar insolation as input. Among the Global horizontal irradiance and Direct horizontal irradiance, which one of them has to be used?
You must use both of them. The PV system is likely inclined. Thus, the irradiation must be computed over this inclined surface. This is done by separate techniques for the direct and the diffuse. The diffuse horizontal is computed from the difference between global and direct. The global irradiance is the sum of these three components: direct B, diffuse D and reflected R:
E(b, a, t) = B(b, a, t) + D(b, a, t) + R(b, a, t)
where b is the tilt angle, a is the azimut angle and t is the time. The irradiance on the inclined plane B(b, a, t) is given by:
B(b, a, t) = B(0, 0, t) cos(q) / cos(qS)
where qS is the solar zenith angle and q is the incidence angle of the sun rays on the inclined surface.
For the diffuse, a very simple model consists in considering the sky vault as a source with uniform radiance. I do not recommend it as this model is inaccurate, but its presentation may improve understanding of diffuse irradiation as you seem to be a beginner in this domain. In this case, as for reflected irradiation, the fraction of sky viewed by the plane may be considered as:
r(b) = (1+ cos(b)) / 2
and D(b, a) = D r(b)
The reflected part is R(b, a) = rg G r(b), where rg is the albedo of the ground, usually set to 0.2 when no information is available.
More complex algorithms are available for computing the diffuse on an inclined plane.
More info at http://www.soda-pro.com/help/general/plane-orientations-and-radiation-components
and at http://www.soda-pro.com/help/helioclim/decomposition-models
You will need GHI, DHI and Temperature. See the following article:
Hussein A. Kazem, Tamer Khatib, and K. Sopian, “Sizing of a standalone photovoltaic/ battery system at minimum cost for remote housing electrification in Sohar, Oman”, Elsevier-Energy and Building, Netherlands, 2013, Vol. 6C, pp. 108-115.