First of all, the Poisson-Boltzmann equation is not, by default, implemented in vanilla NAMD, so MM-PBSA using only NAMD is out of the question. There is an iAPBS/NAMD interface that links NAMD with the APBS (Adaptive Poisson-Boltzmann Solver) method, allowing you, supposedly, to run simulations in Poisson-Boltzmann implicit solvent, but it requires editing both APBS's and NAMD's source codes and recompiling. Plus, it is very buggy (personally, I've never been able to make it work, not even once, even with their tutorials). Anyway, it's on the following website. If you opt to try it, good luck:

http://mccammon.ucsd.edu/iapbs/doc/namd.html

Despite that, NAMD fully implements the Generalized Born method for implicit solvation. Therefore, you can prepare a GB implicit solvation simulation easily, using CHARMM parameters.  In fact, the basic NAMD tutorial has a section on just that:

http://www.ks.uiuc.edu/Training/Tutorials/namd/namd-tutorial-unix-html/index.html

Afterwards, you can use the MM-GBSA method on your results, which is essentially the same idea as MM-PBSA, but with the Generalized Born model.

Or, you can use some other MMPBSA software/suite to analyze your trajectory, such as the MMPBSA.py script in AmberTools 14 (Amber might not be free, but AmberTools is), provided you can convert your files to Amber format, or the g_mmpbsa extension for GROMACS which, supposedly, reads the NAMD DCD file format, provided you have some experience with GROMACS. Have in mind that g_mmpbsa by default does not calculate the entropy part; you can use Carma to do that (http://utopia.duth.gr/~glykos/carma.html).

Note that the tools I mentioned can also read your explicit solvent simulation as well. If you edit your results so that they contain only your protein and ligands, i.e. remove the explicit solvent and any ions you used (You can edit your trajectory using tools such as CatDCD, or with some creative tcl scripting in VMD), you can use these to perform MMPBSA calculations on your results. 

However, if you don't want to use Amber or GROMACS and keep it just with NAMD, it is possible. You just need some software to calculate PB and entropy. Check the following paper for a MM-PBSA calculation on NAMD results (a protein - tRNA complex in explicit solvent): http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3232055/

The authors basically did the following: they calculated the MM part (vdw and coulomb elec) using NAMD itself. That's the easy part.  Afterwards, they used APBS to calculate the solvation energy part (that's the hard part, since it requires a bit of scripting, and it is also computationally demanding). Finally, they estimated the change in entropy using Carma.

I would suggest you first try the VMD and NAMD tutorials to get familiar with their use, or post a question on the NAMD and VMD mailing lists. Also, visit the APBS website and mailing list to find out more on how to use them.  

First of all, the Poisson-Boltzmann equation is not, by default, implemented in vanilla NAMD, so MM-PBSA using only NAMD is out of the question. There is an iAPBS/NAMD interface that links NAMD with the APBS (Adaptive Poisson-Boltzmann Solver) method, allowing you, supposedly, to run simulations in Poisson-Boltzmann implicit solvent, but it requires editing both APBS's and NAMD's source codes and recompiling. Plus, it is very buggy (personally, I've never been able to make it work, not even once, even with their tutorials). Anyway, it's on the following website. If you opt to try it, good luck:

http://mccammon.ucsd.edu/iapbs/doc/namd.html

Despite that, NAMD fully implements the Generalized Born method for implicit solvation. Therefore, you can prepare a GB implicit solvation simulation easily, using CHARMM parameters.  In fact, the basic NAMD tutorial has a section on just that:

http://www.ks.uiuc.edu/Training/Tutorials/namd/namd-tutorial-unix-html/index.html

Afterwards, you can use the MM-GBSA method on your results, which is essentially the same idea as MM-PBSA, but with the Generalized Born model.

Or, you can use some other MMPBSA software/suite to analyze your trajectory, such as the MMPBSA.py script in AmberTools 14 (Amber might not be free, but AmberTools is), provided you can convert your files to Amber format, or the g_mmpbsa extension for GROMACS which, supposedly, reads the NAMD DCD file format, provided you have some experience with GROMACS. Have in mind that g_mmpbsa by default does not calculate the entropy part; you can use Carma to do that (http://utopia.duth.gr/~glykos/carma.html).

Note that the tools I mentioned can also read your explicit solvent simulation as well. If you edit your results so that they contain only your protein and ligands, i.e. remove the explicit solvent and any ions you used (You can edit your trajectory using tools such as CatDCD, or with some creative tcl scripting in VMD), you can use these to perform MMPBSA calculations on your results. 

However, if you don't want to use Amber or GROMACS and keep it just with NAMD, it is possible. You just need some software to calculate PB and entropy. Check the following paper for a MM-PBSA calculation on NAMD results (a protein - tRNA complex in explicit solvent): http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3232055/

The authors basically did the following: they calculated the MM part (vdw and coulomb elec) using NAMD itself. That's the easy part.  Afterwards, they used APBS to calculate the solvation energy part (that's the hard part, since it requires a bit of scripting, and it is also computationally demanding). Finally, they estimated the change in entropy using Carma.

I would suggest you first try the VMD and NAMD tutorials to get familiar with their use, or post a question on the NAMD and VMD mailing lists. Also, visit the APBS website and mailing list to find out more on how to use them.  

Thank you Fotis Baltoumas for your valuable information.

Firstly, i will try with NAMD.If its not possible then, i wil try with other alternatives you mentioned.I wil get back to you if i have any problem.

In implicit simulation of protein.after generating psf, can we directly go to minimization..? without performing solvation and ionization.my view is, they are not required since we are doing implicit solvent simulation.so,are there any other steps to follow after psf generation.

how do we compensate the charge distribution on protein(in case of charged protein),since we are not performing ionization step.?