Effects of Dynamic Penalty Parameters on the Convergence of MOGA in Optimization of a Large Gas Pipeline Network