基于混合深度学习方法的分数阶PID-PSS设计用于抑制电力系统振荡

Steep increase in power demand led to structural change of conventional power networks. Modern power systems include advanced devices and equipment that make it challenging to maintain a reliable and secure power supply. The presence of low frequency oscillation (LFO) is a prominent phenomenon in modern power systems. Proper suppression of these oscillations is required to prevent rotor angle instability. Power system stabilizers (PSSs) are generally employed to tackle this issue. However, conventional PSSs are not capable of damping LFOs effectively in modern networks. Consequently, a fractional-order proportional integral derivative (FO-PID) controller combined with conventional PSS is designed using hybrid deep learning approach in this article. Convolutional neural network (CNN) and long-short term memory (LSTM) are integrated together to form the CNN-LSTM network, which is used to predict the parameters of FO-PID-PSS. Sensitivity analysis is performed to obtain the required hyperparameters of the CNN-LSTM network to be tuned through the brown-bear optimization algorithm (BOA) for enhanced performance. The training of a tuned CNN-LSTM network is done by the actual parameters of FO-PID-PSS obtained through the phase compensation technique (PCT). Diverse test cases of systems operating under contingent conditions are considered to validate the performance of the proposed FO-PID-PSS.