基于小波神经网络无速度传感器DTC系统参数辨识
|
摘要
直接转矩控制(DTC)是20世纪70年代提出的，是继矢量变换控制技术之后，在80年代中期发展起来的一种新型的高性能的交流调速控制技术。目前该技术已成为交流调速传动中的一个热点，本文正是针对这一技术进行了一些较为深入地研究。
为了实现异步电动机高性能的无速度传感器直接转矩控制，必须准确知道电机转速和定子电阻等参数，基于这一点，本文提出了用小波神经网络(WNN)的方法，来构造对定子电阻和转速进行辨识的小波神经网络系统参数辨识器，阐述了小波神经网络的结构并且对小波神经网络的算法，即梯度下降法、基于正交最小二乘法和在线辨识的递推正交二乘法等方法进行了认真的比较、分析
本文从异步电动机模型出发，构造了具有小波神经网络辨识器的直接转矩系统的仿真模型，从仿真结果可以看出，采用小波神经网络的转速辩识器代替DTC系统中的速度传感器，实现DTC系统无速度传感器的方案是可行的、有效的；同时，加入了小波神经网络的定子电阻辨识器的直接转矩控制系统其低速性能还有明显的改善。
本文对无速度传感器直接转矩控制系统的接口电路，包括实现小波神经网络辨识器算法的数字信号处理器(DSP)系统开发的硬件电路，以及对六边形磁链轨迹控制PWM方法和直接转矩控制方案进行了设计和研究。
最后有理由确信，小波变换与神经网络等智能技术的相互交叉、相互渗透，会给无速度传感器直接转矩控制系统的最终稳定运行和准确控制奠定可靠的基础。
one method of AC induction motor control called Direct Torque Control (DTC),which was raised in 1970's .has been developed rapidly in the middle of 1980's and is a powerful control method for motor drives. Following the vector control technique, the quick development of DTC makes itself a new high performance AC driving technique paralleling to the vector control. Now the technique in AC driving is studied widely. In this thesis the author made a study of technique.
To fulfill high performance direct torque control of an induction motor, it is necessary to know precisely parameters of induction motor, such as rotor speed and stator resistance. So wavelets and neural network are combined and wavelet neural network (WNN) is proposed. Through WNN identifier, stator resistance and rotor speed are identified. In this paper, wavelet neural network configuration is expatiated and its algorithm is analyzed. At last conjugate gradient method and orthogonal least squares algorithm and on-linear identification recursive orthogonal least squares algorithm are proposed.
According to induction motor principle, simulation model of direct torque control system with wavelet neural network identifier is built. Simulation result reveals that DTC system with WNN identifier showed excellent static and dynamic performance. So through using speed identifier with WNN replacing speed sensor on the DTC system, the scheme with no speed sensor is effective and feasible, and direct torque control system with WNN stator resistance detector resolved low speed performance
Finally. We developed a hardware system based on Digital Signal Processor (DSP), and to the system can run steadily, we designed the scheme of direct torque control and study the PWM method controlled by the hexagon flux trace.
According to this paper, we can believe that the combination of wavelet transform and neural network will advance the development of intelligent control science and detection technology.
为了实现异步电动机高性能的无速度传感器直接转矩控制，必须准确知道电机转速和定子电阻等参数，基于这一点，本文提出了用小波神经网络(WNN)的方法，来构造对定子电阻和转速进行辨识的小波神经网络系统参数辨识器，阐述了小波神经网络的结构并且对小波神经网络的算法，即梯度下降法、基于正交最小二乘法和在线辨识的递推正交二乘法等方法进行了认真的比较、分析
本文从异步电动机模型出发，构造了具有小波神经网络辨识器的直接转矩系统的仿真模型，从仿真结果可以看出，采用小波神经网络的转速辩识器代替DTC系统中的速度传感器，实现DTC系统无速度传感器的方案是可行的、有效的；同时，加入了小波神经网络的定子电阻辨识器的直接转矩控制系统其低速性能还有明显的改善。
本文对无速度传感器直接转矩控制系统的接口电路，包括实现小波神经网络辨识器算法的数字信号处理器(DSP)系统开发的硬件电路，以及对六边形磁链轨迹控制PWM方法和直接转矩控制方案进行了设计和研究。
最后有理由确信，小波变换与神经网络等智能技术的相互交叉、相互渗透，会给无速度传感器直接转矩控制系统的最终稳定运行和准确控制奠定可靠的基础。
one method of AC induction motor control called Direct Torque Control (DTC),which was raised in 1970's .has been developed rapidly in the middle of 1980's and is a powerful control method for motor drives. Following the vector control technique, the quick development of DTC makes itself a new high performance AC driving technique paralleling to the vector control. Now the technique in AC driving is studied widely. In this thesis the author made a study of technique.
To fulfill high performance direct torque control of an induction motor, it is necessary to know precisely parameters of induction motor, such as rotor speed and stator resistance. So wavelets and neural network are combined and wavelet neural network (WNN) is proposed. Through WNN identifier, stator resistance and rotor speed are identified. In this paper, wavelet neural network configuration is expatiated and its algorithm is analyzed. At last conjugate gradient method and orthogonal least squares algorithm and on-linear identification recursive orthogonal least squares algorithm are proposed.
According to induction motor principle, simulation model of direct torque control system with wavelet neural network identifier is built. Simulation result reveals that DTC system with WNN identifier showed excellent static and dynamic performance. So through using speed identifier with WNN replacing speed sensor on the DTC system, the scheme with no speed sensor is effective and feasible, and direct torque control system with WNN stator resistance detector resolved low speed performance
Finally. We developed a hardware system based on Digital Signal Processor (DSP), and to the system can run steadily, we designed the scheme of direct torque control and study the PWM method controlled by the hexagon flux trace.
According to this paper, we can believe that the combination of wavelet transform and neural network will advance the development of intelligent control science and detection technology.
引文
[1] 杨胤铎，李汉强．运用现代控制理论的异步电机矢量控制技术．武汉交通科技大学学报．2000年04期
[2] Joachim Holtz. Sensorless position control of induction Motors-an emerging technology. IEEE Trans. Ind. Elec. vol.45, No.6. 1998
[3] Kevin D. Hurst, and Thomas G. Habetler. A comparison of spectrum estimation techniques for sensorless speed detection in induction machines. IEEE Trans. Ind. Appl. Vol.33, NO.5,1997
[4] Colin Schauder. Adaptive speed identification for vector control of induction motors without rotational transducers. Conf. Rec. IEEE-IAS annual Meting. 1989
[5] Hirokazu Tajima, and Yoichi Hori. Speed sensorless field-orientation control of the induction machine. IEEE Trans. Ind. Appl. Vol.29, No. 1.1993
[6] Hisao Kubota, Kouki Matsuse, and Takayoshi Nakano. DSP-Based speed adaptive flux observer of induction motor. IEEE Trans. Ind. Appl. Vol.29,NO.2.1993
[7] Fand-Zheng Peng, and Tadashi Fukao. Robust speed identification for speed-sensorless vector control of induction motors. IEEE Trans. Ind. Appl. Vol.30, NO.5. 1994
[8] M. Godoy Simoes. And Bimal K.Bose. Neural network based estimation of feedback signals for a vector controlled induction motor drive. IEEE trans. Ind. Appl. Vol.31, NO.3. 1995
[9] Lazhar Ben-Brahim, Susumu Tasakuma, and Alper Akdag. Speed control of induction motor without rotational transducers. IEEE Trans.Ind. Appl. Vol.35, NO.4.1999
[10] Abdellfattah Ba-Razzouk, Ahmed Cheriti, and Pierre Sicard. Field-oriented control of induction motors using neural-network decouplers. IEEE Trans. Power Elec. Vol.12, NO.4.1997
[11] 吴建兵，刘国海．BP神经网络辩识感应电机转子磁链和转速．电力电子技术．VOL．36．2002
[12] 陈哲，冯天谨．小波分析与神经网络结构的研究进展．电子科学学刊．第22卷．第3期．2000，5
[13] SzuHarold H, TelferBrian. Neural network adaptive wavelets for signal representation and classification. Op-tical Engineering, 1992, 31(9)
[14] DanbechissI. The wavelet transform, fine frequency localization and signal analysis. IEEE Trans Information Theory,1990,36: 961～1005
[15] ZhangJun, WalterG, MiaoY, etal. Wavelet neural networks for function learning. [J] IEEE Trans on SP,1995,43(6): 1485～1497
[16] Lei Liu. Xueguangshao, wensheng cai, YangJiang, Zhongxiao Pan. Prediction of the Atomic Electron Affinities Using a Wavelet Neural Network
[17] Zhang Q,Benveniste A.Wavelet networks. [J] IEEE Trans on NN,1992, 3(11): 889～898
[18] Szu H, Telfer B and Kadambe S. Neural network adaptive wavelets for signal representation and classification. [J] Optical Engineering, 1992, 36(9): 1907～1916
[19] Baskshi B R, Stephanopoulous G. Wave-net: a multiresolution, hierarchical neural network with localized learning.[J] American Institute Chemical Engineer Journal, 1993, 39(1): 57～81
[20] 张帮礼，李银国，曹长修．小波神经网络及其结构设计方法．模式识别与人工智能．1997．10(3)：197-205
[21] 姚骏，彭汉州，甘强．基于离散小波变换的小波网络学习算法及其在心电信号识别重的应用．CCNS．南京，1997，2：605-608
[22] 李向武等．小波神经网络的动态系统辨识方法及应用．控制理论与应用．1998．15(4)：494-500
[23] 金龙，罗莹等．基于小波神经网络的长期预报研究．CCNS，南京．1997，2：617-620
[24] 一种用于函数学习的小波神经网络．自动化学报，1998，24(4)：548-551
[25] 吕柏权等．基于BP神经网络与小波控制研究．系统仿真学报．1997，9(1)：40-48
[26] 马宪民．异步电动机转子转速的非线性辩识与估计．西安科技学院学报．Vol．20 NO．1．2000
[27] 曹承志，曲红梅，路战红，杨晓波．MATLAB软件包中SIMULINK环境下直接转矩控制系统的仿真．电机与控制学报，2001，6，5(2)．111～114
[28] 贾建强，韩如成，左龙．基于MATLAB／SIMUIANK的交流电机调速系统建模与仿真，电机与控制学报，2000，6，4(2)：91～93．
[29] 马宪民，桂佑林．基于DSP的直接转矩控制系统的设计与实现，自动化与仪器仪表，2002(2)：44～45．
[30] 施保华，秦娟英．基于MATLAB／SIMULINK动态仿真模型的异步电动机直接转矩控制的研究，电气传动自动化，2001，12，23(6)．
[31] 曹承志编著．微型计算机控制新技术．北京：机械工业出版社，2001．
[32] 孙亚飞，王旭，贾芙蓉．基于人工神经网络的感应电机系统辨识．控制工程．vol9，No4．2003 4
[33] 薛定宇，陈阳泉编．基于MATLAB／SIMULINK的系统仿真技术与应用．北京：清华大学出版社，2002．
[34] 李夙编．异步电动机直接转矩控制．北京：机械工业业出版社，1998．
[35] James N Nash. Direct Torque Control, Induction Motor Vector Control Without an Encoder. IEEE Trans. On Industry Application, 1997, 33(2): 333～341.
[36] 王海仙，陈福民．感应电动机直接转矩控制系统的数字仿真，电工技术学报，1995，8(3)：6～10．
[37] 喻宗泉．智能控制技术进展，自动化与仪器仪表，2000(1)：2～3．
[38] 张桂新．异步电动机直接转矩控制系统的建模与仿真．中小型电机．2001，28(1)
[39] I.Takahashi, Y.Ohmori. High-performance Direct Torque Control of an Induction Motor. IEEE Trans On IA, 1989(2): 257～264
[40] 聂春燕．直接转矩控制的异步电动机动态仿真模型．微特电机．2001年第4期
[41] 黄海斌，黄显林，胡恒章．基于小波神经网络的多节点机械手滑模自适应控制．电气传动自动化．第22卷，第2期．2000，4
[42] ShyhjierH, etal. Sensorless speed identification of vector-control led induction motor via neural-network based estimation[J]. Electric Power Systems Resear-ch, 1998, 48: 1～10
[43] MarinoP,etal.Linear quadratic state feedback and robust neural network estimat or for field-oriented-control led induction motors[J]. IEEE Trans Industr-ial Electronics,1999,IE46(1): 150～161
[44] 张红英 吴斌．小波神经网络的研究及其展望．西南工学院学报．第17卷 第1期2002年3月
[45] 钱峻，邵惠鹤．小波神经网络结构设计新算法．上海交通大学学报．第33卷第4期 1999年4月
[46] 张帮礼．小波神经网络的构造及其算法的鲁棒性分析．重庆大学学报．vol，18(6)，1995，11
[47] 庞晓虹．三相异步电动机直接转矩控制的一种实现方法．新疆石油学院学报．第13卷第4期2001年12月
[48] 刘泓 莫玉龙．小波神经网络逼近非线性函数算法．应用科学学报．第16卷第2期1998年6月
[49] 沈雪勤，丁宇新．小波变换与小波神经元的选择．河北工业大学学报．vol 25，1996，2
[50] 刘建春，王正欧．一种小波神经网络的快速学习算法及其应用．天津大学学报．第34卷 第4期2001年7月
[51] Liu, Li J P, Xiong J H, etal. The compression of IR Spectra by using waveliet neural network[J]. Chinese Science Bulletin, 1997, 17(2): 822～825
[2] Joachim Holtz. Sensorless position control of induction Motors-an emerging technology. IEEE Trans. Ind. Elec. vol.45, No.6. 1998
[3] Kevin D. Hurst, and Thomas G. Habetler. A comparison of spectrum estimation techniques for sensorless speed detection in induction machines. IEEE Trans. Ind. Appl. Vol.33, NO.5,1997
[4] Colin Schauder. Adaptive speed identification for vector control of induction motors without rotational transducers. Conf. Rec. IEEE-IAS annual Meting. 1989
[5] Hirokazu Tajima, and Yoichi Hori. Speed sensorless field-orientation control of the induction machine. IEEE Trans. Ind. Appl. Vol.29, No. 1.1993
[6] Hisao Kubota, Kouki Matsuse, and Takayoshi Nakano. DSP-Based speed adaptive flux observer of induction motor. IEEE Trans. Ind. Appl. Vol.29,NO.2.1993
[7] Fand-Zheng Peng, and Tadashi Fukao. Robust speed identification for speed-sensorless vector control of induction motors. IEEE Trans. Ind. Appl. Vol.30, NO.5. 1994
[8] M. Godoy Simoes. And Bimal K.Bose. Neural network based estimation of feedback signals for a vector controlled induction motor drive. IEEE trans. Ind. Appl. Vol.31, NO.3. 1995
[9] Lazhar Ben-Brahim, Susumu Tasakuma, and Alper Akdag. Speed control of induction motor without rotational transducers. IEEE Trans.Ind. Appl. Vol.35, NO.4.1999
[10] Abdellfattah Ba-Razzouk, Ahmed Cheriti, and Pierre Sicard. Field-oriented control of induction motors using neural-network decouplers. IEEE Trans. Power Elec. Vol.12, NO.4.1997
[11] 吴建兵，刘国海．BP神经网络辩识感应电机转子磁链和转速．电力电子技术．VOL．36．2002
[12] 陈哲，冯天谨．小波分析与神经网络结构的研究进展．电子科学学刊．第22卷．第3期．2000，5
[13] SzuHarold H, TelferBrian. Neural network adaptive wavelets for signal representation and classification. Op-tical Engineering, 1992, 31(9)
[14] DanbechissI. The wavelet transform, fine frequency localization and signal analysis. IEEE Trans Information Theory,1990,36: 961～1005
[15] ZhangJun, WalterG, MiaoY, etal. Wavelet neural networks for function learning. [J] IEEE Trans on SP,1995,43(6): 1485～1497
[16] Lei Liu. Xueguangshao, wensheng cai, YangJiang, Zhongxiao Pan. Prediction of the Atomic Electron Affinities Using a Wavelet Neural Network
[17] Zhang Q,Benveniste A.Wavelet networks. [J] IEEE Trans on NN,1992, 3(11): 889～898
[18] Szu H, Telfer B and Kadambe S. Neural network adaptive wavelets for signal representation and classification. [J] Optical Engineering, 1992, 36(9): 1907～1916
[19] Baskshi B R, Stephanopoulous G. Wave-net: a multiresolution, hierarchical neural network with localized learning.[J] American Institute Chemical Engineer Journal, 1993, 39(1): 57～81
[20] 张帮礼，李银国，曹长修．小波神经网络及其结构设计方法．模式识别与人工智能．1997．10(3)：197-205
[21] 姚骏，彭汉州，甘强．基于离散小波变换的小波网络学习算法及其在心电信号识别重的应用．CCNS．南京，1997，2：605-608
[22] 李向武等．小波神经网络的动态系统辨识方法及应用．控制理论与应用．1998．15(4)：494-500
[23] 金龙，罗莹等．基于小波神经网络的长期预报研究．CCNS，南京．1997，2：617-620
[24] 一种用于函数学习的小波神经网络．自动化学报，1998，24(4)：548-551
[25] 吕柏权等．基于BP神经网络与小波控制研究．系统仿真学报．1997，9(1)：40-48
[26] 马宪民．异步电动机转子转速的非线性辩识与估计．西安科技学院学报．Vol．20 NO．1．2000
[27] 曹承志，曲红梅，路战红，杨晓波．MATLAB软件包中SIMULINK环境下直接转矩控制系统的仿真．电机与控制学报，2001，6，5(2)．111～114
[28] 贾建强，韩如成，左龙．基于MATLAB／SIMUIANK的交流电机调速系统建模与仿真，电机与控制学报，2000，6，4(2)：91～93．
[29] 马宪民，桂佑林．基于DSP的直接转矩控制系统的设计与实现，自动化与仪器仪表，2002(2)：44～45．
[30] 施保华，秦娟英．基于MATLAB／SIMULINK动态仿真模型的异步电动机直接转矩控制的研究，电气传动自动化，2001，12，23(6)．
[31] 曹承志编著．微型计算机控制新技术．北京：机械工业出版社，2001．
[32] 孙亚飞，王旭，贾芙蓉．基于人工神经网络的感应电机系统辨识．控制工程．vol9，No4．2003 4
[33] 薛定宇，陈阳泉编．基于MATLAB／SIMULINK的系统仿真技术与应用．北京：清华大学出版社，2002．
[34] 李夙编．异步电动机直接转矩控制．北京：机械工业业出版社，1998．
[35] James N Nash. Direct Torque Control, Induction Motor Vector Control Without an Encoder. IEEE Trans. On Industry Application, 1997, 33(2): 333～341.
[36] 王海仙，陈福民．感应电动机直接转矩控制系统的数字仿真，电工技术学报，1995，8(3)：6～10．
[37] 喻宗泉．智能控制技术进展，自动化与仪器仪表，2000(1)：2～3．
[38] 张桂新．异步电动机直接转矩控制系统的建模与仿真．中小型电机．2001，28(1)
[39] I.Takahashi, Y.Ohmori. High-performance Direct Torque Control of an Induction Motor. IEEE Trans On IA, 1989(2): 257～264
[40] 聂春燕．直接转矩控制的异步电动机动态仿真模型．微特电机．2001年第4期
[41] 黄海斌，黄显林，胡恒章．基于小波神经网络的多节点机械手滑模自适应控制．电气传动自动化．第22卷，第2期．2000，4
[42] ShyhjierH, etal. Sensorless speed identification of vector-control led induction motor via neural-network based estimation[J]. Electric Power Systems Resear-ch, 1998, 48: 1～10
[43] MarinoP,etal.Linear quadratic state feedback and robust neural network estimat or for field-oriented-control led induction motors[J]. IEEE Trans Industr-ial Electronics,1999,IE46(1): 150～161
[44] 张红英 吴斌．小波神经网络的研究及其展望．西南工学院学报．第17卷 第1期2002年3月
[45] 钱峻，邵惠鹤．小波神经网络结构设计新算法．上海交通大学学报．第33卷第4期 1999年4月
[46] 张帮礼．小波神经网络的构造及其算法的鲁棒性分析．重庆大学学报．vol，18(6)，1995，11
[47] 庞晓虹．三相异步电动机直接转矩控制的一种实现方法．新疆石油学院学报．第13卷第4期2001年12月
[48] 刘泓 莫玉龙．小波神经网络逼近非线性函数算法．应用科学学报．第16卷第2期1998年6月
[49] 沈雪勤，丁宇新．小波变换与小波神经元的选择．河北工业大学学报．vol 25，1996，2
[50] 刘建春，王正欧．一种小波神经网络的快速学习算法及其应用．天津大学学报．第34卷 第4期2001年7月
[51] Liu, Li J P, Xiong J H, etal. The compression of IR Spectra by using waveliet neural network[J]. Chinese Science Bulletin, 1997, 17(2): 822～825