CAJ | 학술논문

针对传统滑模变结构控制在三相电压型 PWM 整流器中应用时参数摄动所引起的抖动现象，提出一种改进 PID 神经网络的滑模变结构在线控制方法，将 PID 三个参数作为神经网络隐藏层的神经元，利用 PID 算法响应快、无静差的特点以及神经网络的在线自学习能力，实时对滑模趋近律参数进行修改，从而缩短系统状态进入滑模面的时间并减小抖动。对选取的价值函数进行改进，使算法不会陷入局部最优而逼近全局最优解，并对系统的全局稳定性进行分析。通过仿真和实验验证，结果表明该方法能使系统全局稳定，抖动有明显削弱且具有更好的动态响应。
침대전통활모변결구공제재삼상전압형 PWM 정류기중응용시삼수섭동소인기적두동현상，제출일충개진 PID 신경망락적활모변결구재선공제방법，장 PID 삼개삼수작위신경망락은장층적신경원，이용 PID 산법향응쾌、무정차적특점이급신경망락적재선자학습능력，실시대활모추근률삼수진행수개，종이축단계통상태진입활모면적시간병감소두동。대선취적개치함수진행개진，사산법불회함입국부최우이핍근전국최우해，병대계통적전국은정성진행분석。통과방진화실험험증，결과표명해방법능사계통전국은정，두동유명현삭약차구유경호적동태향응。
For the problem that the system input parameters exists disturbances when the traditional sliding-mode varia-ble-structure control(SMVSC)is applied to the three-phase voltage source PWM rectifier,an online solution that sliding-mode variable-structure control base on improved PID neural network design was presented in this paper,taking three pa-rameters of PID as neurons of neural network in the hidden layer and considering that PID algorithm is of fast response,no static error and the online self-learning ability of neural network,the PID algorithm and neural network is combined to modi-fy the sliding approaching rate parameter in real-time,thus the time of system state into the sliding surface and jitter is re-duced.Through improving the selected value function,the algorithm cannot fall into local minimum and the global optimal solution is approached;also,the overall stability of the system is analyzed.Finally into simulation and experimental valida-tion research,show that the method possesses smaller shake and preferable dynamic response.