Damping of subsynchronous resonance using self‐tuning PID and wavelet neural network

Abstract

PurposeSubsynchronous resonance (SSR) problem is often created in generator rotor systems with long shafts (non‐rigid shaft) and large inertias constituting a weakly damped mechanical system. When the electrical network resonance frequency (in which the transmission line is compensated by series capacitors) approaches shaft natural frequencies, the electrical system increases torsional torques amplitude on the shaft. The purpose of this paper is to propose a self‐tuning proportional, integral, derivative (PID) controller to damp the SSR oscillations in the power system with series compensated transmission lines.Design/methodology/approachTo accommodate the PID controller in all power system loading conditions, the gradient descent (GD) method and a wavelet neural network (WNN) are used to update the PID gains on‐line. All parameters of the WNN are trained by the gradient descent method using adaptive learning rates (ALRs). The ALRs are derived from discrete Lyapunov stability theorem, which are applied to guarantee the convergence of the proposed control system. Also, the suggested controller is designed based on a non‐linear model.FindingsThe proposed self‐tuning PID controller is applied to a power system non‐linear model. Simulation results are used to demonstrate the effectiveness and performance of the proposed controller. It has been shown that self‐tuning PID is able to damp the SSR under any circumstances, because the WNN ensures the robustness of the controller. Simplicity and practicality of the proposed controller with its excellent performance make it ideal to be implemented in real excitation systems.Originality/valueThe proposed self‐tuning PID approach is interesting for the design of an intelligent control scheme based on non‐linear model to damp the torsional oscillations. In this suggested controller, the system conditions and requirements adjust on‐line the PID gains. On other words, to damp the SSR, PID gains are intelligently computed by the controlled system. The main contributions of this paper are: the overall control system is globally stable and hence, the SSR is controlled; the control error can be reduced to zero by appropriate chosen parameters and learning rates; and the self‐tuning PID can achieve favorable controlling performance.