基于深度神经网络的HL-2A等离子体水平位移研究

doi:10.16568/j.0254-6086.202202015

摘要/Abstract

摘要： 基于门控循环单元(GRU)的神经网络，构建预测模型的网络拓扑结构，训练和测试了HL-2A装置等离子体水平位移系统响应模型。测试结果显示了该模型对43%的样本数据的拟合度超过80%。把该网络模型作为被控对象，使用基于径向基函数(RBF)神经网络的模型参考自适应控制(MRAC)算法，设计了一个HL-2A等离子体水平位移的MRAC系统。仿真结果显示，该控制系统的输出响应能快速地跟踪各种输入参考信号，控制器能够较好地控制等离子体的水平位移并具有强的抗扰动能力。

关键词: HL-2A装置, 等离子体水平位移控制, 位移响应模型, 门控循环单元, 模型参考自适应控制

Abstract: Based on the gated recurrent unit (GRU), the network topology of the prediction model is built, to train and test the response model of the HL-2A plasma horizontal displacement system. The test results show that the fitting degree of 43% of the sample data exceeds 0.8. Using the network model as the controlled object, an HL-2A plasma horizontal displacement MRAC system is designed with a model reference adaptive control (MRAC) algorithm based on radial basis function (RBF) neural network. The simulation results show that the output response of the control system can quickly track various input reference signals. The controller can control the horizontal displacement of the plasma and has strong anti-disturbance capability.

Key words: HL-2A tokamak, Plasma horizontal displacement control, Displacement response model, Gated recurrent unit, Model reference adaptive control

中图分类号:

TL62+9

付贤飞, 杨斌, 王世庆, . 基于深度神经网络的HL-2A等离子体水平位移研究 [J]. 核聚变与等离子体物理, 2022, 42(2): 264-270.

FU Xian-fei, YANG Bin, , WANG Shi-qing, . Modeling and control of HL-2A plasma horizontal displacement based on neural networks [J]. Nuclear Fusion and Plasma Physics, 2022, 42(2): 264-270.

参考文献

[1] Wang Y, Xiao B, Liu L, et al. System identification for EAST plasmashape and position control [J]. Fusion Engineering and Design, 2018, 29: 140−146．

[2] Shafranov V D. Plasma equilibrium in a magnetic field [J]. Reviews of Plasma Physics, 1966, 2: 103.

[3] Sharma A S, Limebeer D J N, Jaimoukha I M, et al. Modeling and control of TCV [J]. IEEE Transactions on Control Systems Technology, 2005, 13(3): 356−369.

[4] Coutlis A, Bandyopadhyay I, Lister J B, et al. Measurement of the open loop plasma equilibrium response in TCV [J]. Nucl. Fusion, 1999, 39(5): 663−683.

[5] Lister J B, Sharma A, Limebeer D J N, et al. Plasma equilibrium response modelling and validation on JT-60U [J]. Nucl. Fusion, 2002, 42(6): 708−724.

[6] Vitelli R, Boncagni L, Mecocci F, et al. An anti-windup-based solution for the low current nonlinearity compensation on the FTU horizontal position controller [C]. IEEE Conference on Decision and Control. IEEE, 2010.

[7] Rasouli H, Rasouli C, Koohi A. Identification and control of plasma vertical position using neural network in Damavand tokamak [J]. Rev. Sci. Instrum., 2013, 84(2): 013503−201.

[8] Hochreiter S, Schmidhuber J. Long short-term memory [J]. Neural Computation, 1997, 9(8): 1735−1780.

[9] Cho K, Van Merrienboer B, Bahdanau D, et al. On the properties of neural machine translation: Encoder-Decoder approaches [J]. Computer Science, 2014, 1: 103−111.

[10] 赵申剑, 黎彧君, 符天凡, 等译. 深度学习 [M]. 北京: 人民邮电出版社, 2017. 13.

[11] Fu R, Zhang Z, Li L. Using LSTM and GRU neural network methods for traffic flow prediction [C]. 2016 31st Youth Academic Annual Conference of Chinese Association of Automation (YAC). IEEE, 2016.

[12] Kim Ha Young, Won Chang Hyun. Forecasting the volatility of stock price index: A hybrid model integrating LSTM with multiple GARCH-type models [J]. Expert Systems with Applications, 2018, 103(1): 25−37.

[13] Dai J., Song H., Sheng G.. Prediction method for power transformer running state based on LSTM network [J]. High Voltage Engineering, 2018, 44(4): 1099−1106.

[14] Julian, Kates-Harbeck, Alexey, et al. Predicting disruptive instabilities in controlled fusion plasmas through deep learning [J]. Nature, 2019, 568: 526−531.

[15] Srivastavan, Hinton G, Krizhevskya, et al. Dropout: A simple way to prevent neural networks from overfitting [J]. Journal of Machine Learning Research, 2014, 15(1): 1929−1958.

[16] Willmottcj. On the validation of models [J]. Physical Geography, 1981, 2(55): 184−194.

[17] 李波, 宋显明, 李立, 等. HL-2A装置等离子体水平位移控制 [J]. 核聚变与等离子体物理, 2004, 24(4): 269−272.

[18] WANG Xueli. Serf-tuning PID control of electric hydraulic constant power speed regulation based on RBF neural network [J]．Machine Tool and Hydraulics, 2016, 44(22): 115−117．

[19] 张绍德. 一类神经网络非线性系统模型参考自适应控制 [J]. 华中理工大学学报, 2000, 28(4): 77−79.

[20] 姜向龙, 程善美, 李叶松, 等. 神经网络模型参考自适应控制算法研究 [J]. 华中科技大学学报(自然科学版), 2003, 031(1): 4−6.

[21] Wang X, Xiang C, Najjaran H, et al. Robust adaptive fault-tolerant control of a tandem coaxial ducted fan aircraft with actuator saturation [J]. Chinese Journal of Aeronautics, 2018, 6: 1298−1310.

[22] Chien Chiang-Ju, Sun King-Chuan, Wu A-Chen, et al. A robust MRAC using variable structure design for multivariable plants [J]. Automatica, 1996, 32(6): 833−848.

[23] 王雪丽．基于RBF神经网络电液恒功率调速自整定PID控制 [J]．机床与液压, 2016, 44(22): 115−117．

[24] Hartman E J, Keeler J D, Kowalski J M. Layered neural networks with Gaussian hidden units as universal approximations [J]. Neural Computation, 1990, 2: 210−215.

[25] Park J, Sandberg I. Universal approximation using radial-basis-function networks [J]. Neural Computation, 2014, 3(2): 246−257.

[26] Kobayashi H, Ozawa R. Adaptive neural network control of tendon-driven mechanisms with elastic tendons [J]. Automatica, 2003, 39(9): 1509−1519.