Genetic algorithms based self-tuning regulator

Genetic algorithms (GAS) have been proven as robust search procedures. Numerous researchers have established the validity of GAS in optimization, machine learning, and control applications. This paper presents a new intelligent regulator with seytuning scheme using the robust search feature of GAS incorporating with the basic idea of self-tuning regulators. The proposed regulator utilizes GAS to search for the changes of system parameters and to calculate the corresponding control law. The optimum parameters and control law are chosen by means ofthe selection mechanism of GAS which employs the squares oj difference between the actual and the estimated outputs as the fitmss function. The regulator has an on-line parameter identipcation function and requires neither prior knowledge nor training data for learning. The proposed GAS-based is applied to the load frequency control of a power system in order to investigate its effectiveness. As demonstrated by the results obtained from computer simulations, the intelligent regulator can provide good system characteristics.