摘要
以恒温空调系统为控制对象,对神经模糊控制器、常规模糊控制器和PID控制器进行了数字仿真,并用单纯形法对控制比例因子进行了参数寻优,获得了最优参数和动态响应曲线;通过对神经模糊控制器的优化学习,大大提高了神经模糊控制器的控制精度和稳定性,其性能优于最优化的PID控制器和最优化的常规模糊控制器,能有效地满足温度控制要求,并具有较好的鲁棒性;由于神经模糊控制器具有模糊控制和神经网络的智能,经过优化学习后,它具有良好的控制性能和自适应能力。
A digital simulation was conducted for a constant temperature air conditioning system with three different controllers, i.e., the neural fuzzy controller, the conventional fuzzy controller and PID （Proportional-Integral- Derivative） controller. The proportional factors of these controllers were optimized using the simplex method and dynamic response profiles were obtained. The neural fuzzy controller was then trained for optimization with the samples from the optimized PID controller. Its control performance and stability was found to be improved significantly and even superior to the optimized PID controller and optimized conventional fuzzy controller. The system temperature was controlled with desired performance. The robustness of the trained neural fuzzy controller is comparable to the optimized PID controller. This demonstrates that the neural fuzzy controller can be trained for optimization to achieve better control performance and self-adaptability since it has the inherent intelligence of both fuzzy control and neural network.
出处
《制冷空调与电力机械》
2007年第3期1-6,共6页
Refrigeration Air Conditioning & Electric Power Machinery
关键词
数字仿真
参数寻优
优化学习
神经模糊控制器
digital simulation
system optimization
training for optimization
neural fuzzy controller
