Proportional-Integral-Derivative and Fractional Order Proportional-Integral-Derivative Controller Design for Load Frequency Control of A Two Area Automatic Generation Control System Tuned by Optimization Algorithms and Artificial Neural Networks

Abstract

This paper describes how to tune Proportional Integral Derivative (PID) and Fractional Order PID (FOPID) controllers for Load Frequency Control (LFC) of two area linked thermal power systems using an optimization technique and Neural Network (NN) prediction-based approaches. Generator speed fluctuates due to frequent variations in load demand. As a result, frequency deviates from the specified value. That is why it is critical to employ a correctly tuned controller to change the generator's speed in order to keep the frequency as near to its rated value as feasible. The Enhanced Gradient Based Optimizer (EGBO) is utilized in this study to adjust a PID controller by optimizing an Integral Time multiplied by Absolute Error (ITAE) based fitness function. The EGBO algorithm's results were compared to the Gradient Based Optimizer (GBO), Chimp Optimization Algorithm (ChOA), Sine Cosine Algorithm (SCA), Grey Wolf Optimization (GWO), and Particle Swarm Optimization algorithms (PSO). When compared to alternative optimization approaches, the relevant data reveal that the EGBO algorithm is competitively superior in terms of robustness, accuracy, and latency. The ITAE-based fitness function of the FOPID controller is also optimized using EGBO. In general, the FOPID controller outperforms the PID controller in terms of ITAE since it has a few more configurable parameters. In order to optimize the ITAE, optimization methods require some time. As a result, Artificial Neural Networks (ANNs) are used for tuning with nearly little latency. Datasets are created using the previously described PID controller optimization procedures, which are then utilized to train different hyperparameters.