Neural network based self-tuning PID controller for automatic voltage regulator of hydropower plant

Abstract

Hydropower plant is a renewable resource with low operating, maintenance expenses and low environmental effects. Due to the constant load change as a result of changing consumers demand, terminal voltage from the generator is fluctuating for certain period before it settle to the desired level. The amount of fluctuation is negatively influencing the power quality and performance of power system. Automatic voltage regulator (AVR) is playing vital role for maintaining the terminal voltage within desired level. The proportional-integral-derivative (PID) controller's is popularly deployed in AVR system due to its ease structure and straightforward design with almost no computational cost. However, traditional methods of PID tuning in some industrial applications does not meet the required response due to severe load fluctuations. In this work, in order to meet the optimum PID-AVR performance; a three types of neural networks namely: feed forward neural network (FFNN), cascade back propagation neural network (CBPNN), and convolutional neural network (CNN) were used to design three self-tuning PID controllers (NNs-PIDF) for AVR system. These artificial intelligence based controller are proven stunning performance over traditional PID controllers also over those controller made using particle swarm optimization (PSO) and fuzzy logic. The outcomes of this work revealed that FFNN-PIDF based AVR system was able to produce best results e.g. settling time, overshoot, and rise time. The proposed controllers has provided consistence performance in controller stability and robustness tests.