In view of the Differential Evolution (DE) optimization framework, the best mutation strategy for solving a particular optimization problem, depends on the problem itself.

In general, the best solution for a problem is determined through trial-end-error because the so-called "no free lunch" theorem suggests that a particular optimization algorithm is suitable for one problem, but unsuitable for another.

The following literature is readily informative on the matter:

1. Wolpert, D.H., Macready, W.G., "No Free Lunch Theorems for Optimization", IEEE Transactions on Evolutionary Computation 1 (1997) 67-82.

2. Ho, Y.-C., Pepyne, D.L., «Simple Explanation of the No Free Lunch Theorem of Optimization», Proc. 40th IEEE Conf. Desicion Contr., December 2001, Orlando FL USA, 4409-4414. (http://www.cc.gatech.edu/~isbell/reading/papers/nfl-optimization-explanation.pdf)

I think you may try some existing efficient DE variants first. 

You can get many codes from this homepage.

http://ist.csu.edu.cn/YongWang.htm

However, the most appropriate mutation strategies and parameters required by DE generally are different  to solve  different optimization problems.

When considering mutation, you must take into consideration a multitude of factors. Below is a fragment from an extensive DE review which I will soon publish:

"At a general level, the most effective strategies (among the initial variants) are Rand/1 and Best/2 [1].The experimental results showed that the strategies based on the best solution found so far have a fast convergence speed and perform well on uni-modal problems, but are more likely to get stuck at local optimum when solving multi-modal problems [2]. They promote the exploitation of the best chromosome strategy which in turn, has an effect on the diversity [3]. Distinctively, the mutation operators based on randomly chosen base vectors enhance exploration, as a high degree of variability affects the mutation [4]."

Reference List

[1] C. HU and X. YAN, Computers & Mathematics with Applications, 62 (2011) 32-43.

[2] A.K. Qin, V.L. Huang, and P.N. Suganthan, Evolutionary Computation, IEEE Transactions on, 13 (2009) 398-417.

[3] J. Arabas, L. Bartnik, and K. Opara, DMEA An algorithm that combines differential mutation with the fitness proportionate selection, Differential Evolution (SDE), 2011 IEEE Symposium on, 2011, pp. 1-8.

[4] M.G. Epitropakis, D.K. Tasoulis, N.G. Pavlidis, V.P. Plagianakos, and M.N. Vrahatis, IEEE Transactions on Evolutionary Computation, 15 (2011) 99-119.

@ Athanasios Kakalis : thanks for the reply and for the references.

@ Guohua Wu: thanks for the link, really helpful

@ Elena Niculina Dragoi: thanks for the response and the passage, looking forward to read your paper.

Similar to optimization techniques. There is no a mutation strategy that solves compeptitively all problems. This thepends on the problema at hand

Althoug the best/ es the most employed and the used by default in several codes