电工技术学报  2024, Vol. 39 Issue (6): 1886-1897    DOI: 10.19595/j.cnki.1000-6753.tces.222367
电力电子 |
基于改进灰色预测单神经元PI的全超导托卡马克核聚变发电装置快控电源电流控制
黄海宏, 陈昭, 王海欣
合肥工业大学电气与自动化工程学院 合肥 230009
Current Control of Experimental Advanced Superconducting Tokamak Fast Control Power Supply Based on Improved Grey Prediction Single Neuron PI
Huang Haihong, Chen Zhao, Wang Haixin
School of Electrical Engineering and Automation Hefei University of Technology Hefei 230009 China
全文: PDF (6163 KB)   HTML
输出: BibTeX | EndNote (RIS)      
摘要 全超导托卡马克核聚变发电装置(EAST)快控电源负载电感的电流受多种不确定环境因素的影响而难以预测,灰色预测无需精确对象模型,只需少量已知信息即可实现输出电流短期预测,已在EAST快控电源中有了一定研究应用。为解决灰色预测在EAST快控电源中对突变信号边沿预测精度低和预测延时时间长的问题,提出一种改进灰色预测算法实现输出电流预测。在一个开关周期内对输出电流进行等时长间隔采样4次作为原始序列,将滚动式采样预测改为逐周期采样预测,在实现灰色预测的过程中不必依赖过去几个周期的历史采样信息,只需本周期的4个原始采样值即可实现输出电流的预测。根据预测电流与参考电流误差自适应调整单神经元比例-积分(PI)控制的输出增益,实现输出电流的快速准确控制。仿真和实验结果表明,在EAST快控电源电流预测过程中所提出的改进灰色预测,对比传统灰色预测具有更高的预测精度和更小的预测延时,改进灰色预测变增益单神经元PI控制能够在减小电流超调的同时加快输出电流响应速度。
服务
把本文推荐给朋友
加入我的书架
加入引用管理器
E-mail Alert
RSS
作者相关文章
黄海宏
陈昭
王海欣
关键词 EAST快控电源改进灰色预测逐周期采样预测变增益单神经元PI    
Abstract:The current of load inductance in the experimental advanced superconducting Tokamak (EAST) fast control power supply is difficult to predict due to the influence of various uncertain environmental factors. An accurate object model is optional in grey prediction.
When the traditional rolling grey prediction is applied to the EAST fast control power supply, there is a specific prediction error at the abrupt signal, and at least 4 cycles of historical current data are needed to realize the current prediction of the next cycle. There are low prediction accuracy and long prediction delay problems for the edge of abrupt signal in the EAST fast control power supply. Therefore, an improved grey prediction algorithm is proposed to predict the output current. The output current is sampled 4 times at equal time intervals in a switching cycle as the original sequence. The rolling sampling prediction is changed to the cycle-by-cycle sampling prediction. Historical sampling information of the past several periods is needed to realize grey prediction, and four original sampling values of the current period are needed to realize the prediction of output current at the expense of the delay of one switching period. The original data sequence is composed of 4 times sampled at equal time intervals in a switching period, which avoids the value with non-exponential law in the original sequence and further improves the accuracy of grey prediction at the abrupt signal. The performance difference between improved grey prediction and traditional grey prediction is analyzed through convergence analysis. The improved grey prediction has a faster convergence speed and smaller prediction errors. Based on the objective function of the current tracking reference current at the next moment, a single neuron PI parameter adaptive adjustment network structure is constructed. The output gain of the single neuron proportional-integral (PI) control is adaptively adjusted according to the error between the predicted current and the reference current, and the PI parameter is adaptively adjusted online to realize the fast and accurate control of the output current.
The simulation results show that the improved grey prediction proposed in the EAST fast control power supply current prediction process has higher prediction accuracy than the traditional grey prediction. The improved grey prediction variable gain single neuron PI control can reduce the current overshoot and accelerate the output current response speed. Compared with the traditional PI control, the output current overshoot is reduced by 3.125%, the control delay and dynamic time in the rising process are reduced by 40 μs, reaching the rising steady state value 0.34ms ahead of time. The control delay and dynamic time during the drop are reduced by 50 μs and 10 μs. The steady-state value of the drop is reached 0.29 ms ahead. According to the theoretical, simulation, and experimental results, the improved grey prediction variable gain single neuron PI control method proposed can improve the prediction accuracy of the EAST fast control power supply current, reduce the prediction delay, suppress the overshoot of output current, and speed up the dynamic response speed of output current.
Key wordsExperimental advanced superconducting Tokamak (EAST) fast control power supply    improved grey prediction    cycle by cycle sampling prediction    variable gain single neuron PI   
收稿日期: 2022-12-23     
PACS: TM93  
  TM917  
基金资助:国家自然科学基金区域创新发展联合基金资助项目(U22A20225)
通讯作者: 陈 昭 男,1996年生,博士研究生,研究方向为新型电能变换技术等。E-mail: chenzhao_0202@163.com   
作者简介: 黄海宏 男,1973年生,教授,博士生导师,研究方向为新型大功率变流技术与电力电子技术等。E-mail: hhaihong741@126.com
引用本文:   
黄海宏, 陈昭, 王海欣. 基于改进灰色预测单神经元PI的全超导托卡马克核聚变发电装置快控电源电流控制[J]. 电工技术学报, 2024, 39(6): 1886-1897. Huang Haihong, Chen Zhao, Wang Haixin. Current Control of Experimental Advanced Superconducting Tokamak Fast Control Power Supply Based on Improved Grey Prediction Single Neuron PI. Transactions of China Electrotechnical Society, 2024, 39(6): 1886-1897.
链接本文:  
https://dgjsxb.ces-transaction.com/CN/10.19595/j.cnki.1000-6753.tces.222367          https://dgjsxb.ces-transaction.com/CN/Y2024/V39/I6/1886