The distributed optimal power flow problem is addressed. No assumptions on the problem cost function, and network topology are needed to solve the optimization problem. A distributed particle swarm optimization algorithm is proposed, based on Deb's rule to handle hard constraints. Moreover, the approach enables to treat a class of distributed optimization problems in which the agents share a common optimization variable. Under mild communication assumptions, agents are only required to know local variables, cost function, and constraints to solve a common optimization problem. A simulation example is provided, based on a 5-bus electric grid.