سال انتشار: ۱۳۹۰
محل انتشار: پنجمین کنفرانس بین المللی پیشرفتهای علوم و تکنولوژی
تعداد صفحات: ۹
A Lari – Faculty of Electrical and Computer Engineering, Noushirvani University
A Khosravi –
A Alfi – Faculty of Electrical and Robotic Engineering, Shahrood University of Technology, Shahrood 36199-95161, Iran.
Conventional PID controller is the most popular controller in various field of industry. In spite of strong ability, this controller cannot be usually used for long time delay systems. There is a kind of neural networks called PID neural network (PIDNN) that utilizes the advantages of both PID controller and neural network simultaneously. PIDNN’s weights were in first adjusted by back propagation (BP) algorithm. BP algorithm ensures the final convergence, but the critical drawback is that the study training costs a long time. The convergence into local extremum is also possible. Hence, another method must be opted to optimize the network’s weight. This paper proposes a novel PIDNN without saturate surface namely PIDNN-PSO which its weights are adjusted using particle swarm optimization (PSO). PSO algorithm is an evolutionary optimization algorithm that due to the ease of implementation and fast convergence speed has been widely applied in many areas. The proposed controller is utilized as a controller for long time delay systems. The performance of the proposed controller is compared with the controller is designed by PIDNN-BP algorithm. Simulation results demonstrate the effectiveness of the proposed algorithm.