Particle swarm optimization is best  for control variables ( capacitor, OLTC, and voltage regulator) in a distribution system.

Dear Monsef Tahir,

In fact, according to the “no free lunch” theorem, there is no specific optimization

algorithm to achieve the best performing for all optimization problems. However, in my opinion, PSO algorithm is more computationally efficient and suitable than GA to solve mixed integer nonlinear programming problem. PSO is two populations - pbests and current positions. This allows greater diversity and exploration over a single population (which with elitism would only be a population of pbests). Furthermore, the momentum effects on particle movement can allow faster convergence (e.g. when a particle is moving in the direction of a gradient) and more variety/diversity in search trajectories.

Regards,

Soheila

Dear Monsef,

Certainly PSO has become a must especially for the distribution system, however for complex systems combining with other methods such GA is obtained more performance. For this i suggest you documentation represented by the links and the attached files on the topic.

- A Particle Swarm Optimization for Reactive Power and Voltage Control ...

https://www.researchgate.net/.../220742055_A_Particle_Swarm_Optimizati...

- Ahmad Reza Malekpour | LinkedIn

https://www.linkedin.com/.../ahmad-reza-malekpour-...

-Coordinated Volt/Var Control in Distribution Systems with Distributed ...

www.mdpi.com/2076-3417/6/1/4/pdf

Best regards

From my experience GA has "revolutionary" and PSO "evolutionary" approach in improving objective function. I mean that in GA we have big improvement between iterations, but than in few next iterations there is no improvement at all and than again big improvement. Plot of objective function vs iterations is much like steep stairs. PSO has more uniform rate of convergence, and can get closer to optimum. The same plot is much more smooth and predictable. I've tested the algorithms, with good results, for limited number (~30) of individuals/particles in population to limit the time of computations (~5min for single objective function evaluation). GA might behave better with larger population.

Thank you all for the explanation 

Specifically, if we have more than a hundred of variables, which one will be more faster in terms of time and converge?

For both algorithms time of optimization doesn't depend directly on the number of variables. But in the end, in most cases, PSO converges faster than GA and closer to optimum. Remember that the algorithms are heuristic, so depending on the particular run of the optimization you can sometimes get better results from GA, or the algorithms won't converge to reasonable solution at all. If you have enough time, it is good to run the same optimization case few times to see if the results are repeatable. This way you will also get scattering of the optimization solution.  I think that everyone who answered your question, including me, would recommend to use rather PSO algorithm.