Self-adjusting PID control system using a neural network for a binary power plant

Abstract

Proportional–integral–derivative (PID) control systems are typically used in power plants owing to their simple structure and ease of implementation. During industrial power generation, binary power plants using low-grade thermal energy sources experience fluctuations in characteristic values due to the presence of impurities and corrosive components in the hot water used as a heat source. Moreover, the temperature of the hot water depends on environmental conditions. However, fine-tuning the PID controller parameters during the operation of binary power plants is challenging, with unmodeled dynamics and uncertainties in parameters arising from changes in the characteristic values. In this study, a novel neural network-based self-adjusting PID control system is proposed, establishing a design strategy for effective control of binary power plants. A comparative analysis of simulation results from control systems with fixed conventional PID parameters, a back-propagation neural network, and the proposed method demonstrates that the proposed self-adjusting PID control approach effectively operates the investigated binary power plant.