In addition to the answers by Yomna M. Ibrahim and Amad Nori Abdulqudos
a few notes of caution.
First, you need to define grain size more clearly.
Techniques such as wide angle diffraction will yield the "crystal" size of the individual crystalline particle.
On the other hand techniques like dynamic light scattering , moderate resolution TEM or SEM or small angle scattering will yield the "particle" size.
If your particle consists of several individual crystallites, these two "sizes" may differ drastically.
In a first summary, the nanoparticle size is a function of the measuring technique!
A few notes on the techniques mentioned:
-Dynamic light scattering: This is based on the autocorrelation of the interaction of a laser with the particles suspended in solution. It measures the diffusion speed of the objects in solution interpreted via continuum theory to yield the particle size. This does not take into account that all nanoparticles are surrounded by a layer of "hydration" shells, commonly about two to three solvent molecules in thickness. This is a larger object that will diffuse slower, thus in my opinion DSL overestimates the size.
This gives an estimate in the order of magnitude at best! Do not use it for serious considerations.
It is an extrapolation down towards small crystal sizes!
The Scherrer Parameter K depends on the shape of the object, which of course you do not know.
The FWHM of reflection can be widened by additional unknown factors such as inhomogeneous strain, defects like stacking faults etc
The FWHM also depend on the instrument, and this contribution must be deconvoluted from the experimental widths
For small nanoparticles reflections will often overlap creating a wrongfully broad estimate of FWHM.
If you want to use wide angle diffraction (Xray, neutron or electron), the complete powder diffraction pattern must be simulated by building an atomistic model (including defects) and calculating the powder pattern through the Debye-Scattering-Equation. (of course including a convolution by the instrumental resolution function)
-Small angle scattering: This technique uses the lowest 0 to ~ 3 degrees of a powder diffraction pattern and the average overall size can be estimated by a fit with suitable software packages.
-TEM: Be careful that the sample preparation process does not alter the "grain" size of "particles" by for example separating these into individual smaller crystallite sizes.
As TEM provides a projection of the object, you do not "see" the third dimension, which might be very deceitful in the case of plate like nanoparticles. Unless you have a HRTEM (high resolution) you will not necessarily "see" the internal crystal structure and thus you might not know if this is a particle or a nanocrystal!
-SEM is also an alternative, usually easier on the sample preparation, a bit more limited to slightly larger sizes. Again, make sure that all three dimensions are properly measured.
Both TEM and SEM will show you the overall particle size. You might see object terminated by hkl facets, an indication that the object is a single crystallite object. Ensure that the size and size distribution determination used plenty of particles without the bias of a preference for either larger or smaller particles.
The measurements with DLS are used for dynamic particle size determination. The results of this may differ from the measurements with SEM and TEM. You can use the ImageJ program for the measurements with SEM and TEM images. For this purpose, you can select many particles with the clearest image and determine the average diameter.