In the application of PSO for OPF, I have few queries regarding handling of equality/inequality constraints. What I have learned so far is:

1. Constraints which are based on dependent variables can be included in the objective function through penality function approach. This leaves me with the question about constraints which are based on control variables? How to handle these constraints in terms of OPF? and furthremore, in this case, do I have to run the load flow once each particle value is updated or I just have to check the objective function for each particle value?

2. The second method that I have learn is that once PSO gives new updated values of each particle, run the load flow again. In this way, the load flow, at least, will automatically take care of the equality constraint. Now, the question is what if some inequality constraints are violated, then how to handle this scenario?

I would really appreciate if someone can provide me some good research paper which clearly demonstrate how to handle constraints related to the OPF problem which is solved through PSO?