A combined on-line intelligent-identifier and Pole-Shift (PS) feedback controller as an adaptive power system stabilizer (APSS) is presented in this dissertation. Artificial Neural Networks (ANNs) have excellent approximation capabilities and produce impressive results in tracking the dynamic oscillations in power system. The PS-Control is ideal for its robustness and stability conditions. The hybrid system combines their advantages while avoiding their weaknesses.

Two ANN architectures are proposed for identification:​ a simple adaptive linear element (ADALINE) and a radial basis function (RBF) network. The advantages and shortcomings of the two architectures are discussed.

The ANN-Identifier goes through two-stages of learning:​ off-line and on-line training. The off-line training is used to store the ‘a priori’ knowledge. After the off-line training the networks are further trained every sampling period making it an adaptive approach. The ANNs are also linearized at every sampling instant to obtain the system regression coefficients. The Pole-Shift control uses the above regression coefficients to calculate the closed-loop poles. The unstable poles are moved inside the unit circle in the z – plane and the control calculated so as to achieve the desired performance.

Most of the present control techniques involve the optimization of one-step ahead performance index. However, the control signals have a dynamic impact on the future states of the system. Therefore, full multi-scale optimization is desirable. The drawback of such a full-scale optimization is the computational burden attached to it. To overcome this shortcoming, simplifications are proposed by using the “receding horizon” principle and “dynamic control limits”.

A power system dynamics simulation package (PSDSP) is described for the purpose of conducting computer simulations. The PSDSP uses the interprocess communication facilities of UNIX to achieve a loose coupling between the power system components and the APSS. Modular programming is used throughout this development for flexibility and reusability of the code. Simulations are carried out on different configurations of power systems to assess the performance of the APSSs.

After successful simulation studies, the APSSs are further tested on a scaled physical model of a power system. The performance of APSSs suggests significant benefits over conventional PSSs:​ performance improvement and no requirement for parameter tuning.