Article Preview
Top1. Introduction
Small perturbations (like a short circuit in a specific line or a sudden load increase) can generate electromechanical oscillations. These oscillations, if not well damped, can cause mechanical wear in the generators, affect the power transfer in transmission lines and even interrupt the power supply in extreme cases. The most common electromechanical oscillations modes are local modes (with oscillation frequency typically between 0.7 and 2 Hz) and inter area modes (with oscillation frequency generally between 0.1 and 0.7 Hz) (Kundur, 1994).
Depending on several factors (most related to the operation condition, to the topology and to set of system parameters), the natural damping of these modes can be low, which is detrimental to a stable operation of the system (Rogers, 2000). The insertion of PSSs controllers is the most used manner to increase the damping of the system (Kundur et al., 2004). The PSSs have largely used in power systems since they were proposed in (De Mello & Concordia, 1969). These controllers provide damping to the electromechanical oscillations by a stabilizing signal added to automatic voltage regulator of the generator where the PSS is placed. The primarily function of the PSSs is provide damping to electromechanical local modes. However, in some cases, the inter area modes cannot present a satisfactory damp using just PSSs, being necessary the inclusion of FACTS (Flexible Alternating Current Transmission System) with its respective POD.
Even in the most common cases the application of PSS and POD are not a trivial task. The parameters of these controllers must be tuned to provide an increase of damping for multiples oscillations modes which change according with the operation conditions. This work can make the process rather complex. The techniques for tuning power system controllers have been improved from an initial work presented in (Larsen & Swann, 1981), with a variety of approaches. The initial proposed techniques were performed using trial-and-error processes, among them are enlightened transfer function residues analysis (Pagola et al., 1989), selective modal analysis (Pérez-Arriaga et al., 1982) and the induced torque coefficients (Pourbeik et al., 2002). However, these techniques do not guarantee a satisfactory damping of the power system for several operating points. Thus, it is necessary to test the set of parameters tuned for each operating point considered in the design. If for one operating point the damping provide for the set of controllers parameters is not satisfactory, the design must be done again.
Automatic tuning methods are being studied to facilitate the designers work in the tuning of the controllers parameters. Such methods have the main advantage of considering several operating points of the system simultaneously, yielding a robust controller regarding variations in its nominal operating point. Examples of the many approaches that have been proposed are H∞ mixed-sensitivity theory (Chaudhuri & Pal, 2004), regional pole placement using linear matrix inequalities (Werner et al., 2003) and genetic algorithms (GA) (Do Bomfim et al., 2000). These automatic techniques present significant simplification if compared with techniques based in trial-and-error.
However, genetic algorithms have to repeatedly evaluate the objective values for each individual, which is usually a very time consuming process. Fortunately, it is an inherent nature of the genetic algorithms that the evaluation of individuals be conducted independently, and hence the computing time can be effectively accelerated by means of parallel computation (He et al., 2007). Furthermore, an alternative that is being explored is the parallelization of the algorithm for automatic tuning of power system controllers can be seen in (Castoldi et al., 2010). Since modern computers have more than one core, parallel computing is becoming an important ally in the reduction of time of the algorithms. Thus, parallel computing together with an appropriated optimization method can minimize the time convergence for the design.