摘要
针对传统的基于最优消谐波理论的开关点预置逆变器存在开关角固定只能保证局部工作点有最小总谐波含量(THD)及逆变器环节输出电压不可调的缺点,应用了一种新的神经网络控制器。当系统工作时,由在线神经网络实时调节逆变器输出电压的大小;同时各功率管的开关角由另一个神经网络通过对最优脉宽调制(PWM)最优开关角的拟合,直接计算出来,保证在整个参数变化范围内均工作在最优的开关角状态。整个控制系统具有结构简单、反应灵敏、调压精确、输出电能质量高且全数字化的特点。
A harmonic-elimination PWM control method with neural-net control is presented. The SPWM control method bases on the optimal PWM theory, and the optimal switching angles, under which the THD of output voltage is lowest, are given. Then a neural-net (net A) is used to fit each switching angle curve to reduce the calculation on line. The neural-net is very simple and easy to be realized by DSP system. So less data, the neural-net抯 weights, activation transfer function and threshold values, are needed to saved in the control system of inverter. Another neural-net(net B)is used in the inverter as the on-line controller. It is used to regulate the inverter抯 output voltage to be constant. When input voltage or loads change, the neural-net adjusts its weights according to the error between output voltage and given voltage. Then the switching angles are changed by its output. As the neural-net A, net B is also simple and easy to be realized. In fact, we use a simplest neural-net, one-input-one-output linear single hidden neural-net, to form the controller. The stability of the system is proved, and under the control of neural-net, the error of output voltage descends sharply. At last the feasibility of the method is verified.
出处
《电工技术学报》
EI
CSCD
北大核心
2004年第2期98-102,共5页
Transactions of China Electrotechnical Society
