I recommend reading (citeseerx.ist.psu.edu/viewdoc/download) paper since it proposes an improved maximum power point tracking (MPPT) method for the photovoltaic (PV) system using a modified particle swarm optimization (PSO) algorithm.
The mean drawback of PSO algorithm in MPPT (or GMPPT) application, is slow speed of convergence and search space is large due to the fact that lot of time is wasted in visiting states of poor fitness.
Maximum power point tracking using PSO may be slower a bit. However, it will be better than perturb and observe (P&O) method and can be implement for real filed application.
Training the system with specific set of values are required for PSO also the number of power perturbation steps required is more compared to other evolutionary schemes.
As @Youssef Cheddadi said, "the mean drawback of PSO algorithm in MPPT (or GMPPT) application, is slow speed of convergence and search space is large". Some research set some limitation on the searching area and(or) moving direction, which reduce the searching time as well as high tracking efficiency. However, the PSO method loses its its randomness, and then loses its inherent advantages. In other words, the PSO method with these limitations may not able to track some cases (like an extreme situation).
There can be no debate on the fact that like most evolutionary optimisation systems PSO based MPPT will be quite slow. That is, what Youssef suggested above is very much true !
I would like to add however, that the handicap is mainly in the context of real time implementation. If you employ a PSO for offline design of the MPPT control, then there may be alternative routes that become acceptable.
Check out this suggested route, as an example:
Create an exhaustive simulation of the PV unit with realistic loads and disturbances on a suitable platform (say, PSCAD/EMTDC as an example).
For different combos of load and disturbances that you simulate, use the PSO to obtain optimal control parameters (such as gains).
While implementing, programme a suitable lookup table with the optimised parameters, that will select the best parameters based on sensed load and disturbance conditions.
Implement the control (as a lookup table, database, or fuzzy-logic controller) in real time using a suitably fast embedded controller.
There can be many variations to the above, but notice that we have carefully avoided a PSO optimisation in real time !!
the main problems that face MPPT based on Metaheuristic algorithms are convergence time..so it takes sometimes so much time to reach MPP also this needs huge computing efforts which complicates the real-time implementation.
The main advantage of the method is the reduction of the steady- state oscillation (to practically zero) once the maximum power point (MPP) is located. Furthermore, the proposed method has the ability to track the MPP for the extreme environmental condition, e.g., large fluctuations of insolation and partial shading condition.