Dynamic behavioral modeling of nonlinear circuits using a novel recurrent neural network technique

In this paper, a new method called local-global feedback recurrent neural network (LGFRNN) is proposed for dynamic behavioral modeling of nonlinear circuits. The structure of the proposed method is based on recurrent neural network and constructed by time-delayed local and global feedbacks. Adding time-delayed feedbacks has a great impact on the learning capability of previous neural network-based methods. Moreover, time-delayed local feedbacks alleviate the problem of slow convergency of the conventional neural network-based methods in the training phase. The proposed LGFRNN can be trained only by having sampled input-output waveforms of the original circuit without knowing the internal details of the circuit. A training algorithm based on real-time recurrent learning (RTRL) is used to train LGFRNN. After the training phase, the proposed LGFRNN provides accurate macromodel of a nonlinear circuit. The proposed method is more accurate compared with the conventional neural-based models (which do not benefit from time-delayed local-global feedbacks) and also significantly reduces the training time of the conventional models. Moreover, proposed LGFRNN is faster than the existing models in simulation tools. The validity of the proposed method is verified by time-domain modeling of three nonlinear devices including commercial TI's SN74AHCT540 device, five-stage complementary metal-oxide-semiconductor (CMOS) receiver, and commercial TI's LM324 power amplifier.     

Overview of progress in quantum systems control

The development of the theory on quantum systems control in the last 20 years is reviewed in detail. The research on the controllability of quantum systems is first introduced, then the study on the quantum open-loop control methods often used for controlling simple quantum systems is analyzed briefly. The learning control method and the feedback control method are mainly discussed for they are two important methods in quantum systems control and their advantages and disadvantages are presented. According to the trends in quantum systems control development, the paper predicts the future trends of its development and applications. A complete design procedure necessary for the quantum control system is presented. Finally, several vital problems hindering the advancement of quantum control are pointed out. Translated from Chinese Journal of Quantum Electronics, 2003, 20(1): 1–9 [译自:: 量子电子学报]     

基于测量的量子图像识别研究

目前,已有的量子相似度比较算法:1)逐个比较图像对应位置的像素值;2)将两幅图像分别用量子态表示,再将两幅图像进行连接(意味着将两个量子态连接成一个态),再进行相关的量子操作。所提出的比较算法,是在不连接图像的基础上,将图像用量子态表示,进行控制交换(c-Swap)操作,再进行量子测量,根据测量结果判断两幅图像的相似度。将所提的量子相似度比较算法应用到量子手势识别中,实验结果表明所提算法在识别问题上具有可行性。在经典领域中,手势识别的流程比较复杂。而在量子领域中,无需提取手势的颜色、纹理、特征等步骤,直接可以将手势进行二值化表示,再根据所提的图像相似度算法来实现手势识别。     

Analysis of parasitic quantum effects in classical CMOS circuits

In the mesoscopic regime, the MOS device performance is affected by gate‐induced quantization effects leading to a loss of transconductance and threshold voltage shift and gate leakage tunnelling currents degrading the overall device performance. We discuss the expected impact of quantum effects in highly down scaled CMOS circuits. Based on 1‐d numerical simulations for transport in mesoscopic systems, we set up Spice circuit models. The Spice models rebuild the influence of quantum effects; and the functionality of classical circuit concepts can be ‘tested’ in their robustness against these effects. A few circuit examples will be given. Copyright © 2005 John Wiley & Sons, Ltd.     

基于Buck-Boost拓扑的新量子遗传算法 在MPPT技术中的应用

非线性的光伏电池要求在不同光照和温度等工况条件下实时变更电压、电流等负载特性实现MPPT追踪光伏电池最大功率点。针对一般遗传算法用于MPPT最大功率点追踪过程中出现的早熟、收敛慢等不足,提出一种改进的量子遗传算法与扰动法相结合的算法。在电路拓扑结构上采用Buck-Boost电路替代传统的Buck电路或Boost电路,并在Matlab/Simulink下进行了建模仿真。实验结果表明,该算法具有良好的最优点搜索能力和跟踪能力,且控制精度高,同时有效抑制了最大功率点附近的波动,证明了该控制方法的准确性和有效性。     

A new approach for numerical simulation of quantum transport in double-gate SOI

Numerical simulation of nanoscale double-gate SOI (Silicon-on-Insulator) greatly depends on the accurate representation of quantum mechanical effects. These effects include, mainly, the quantum confinement of carriers by gate-oxides in the direction normal to the interfaces, and the quantum transport of carriers along the channel. In a previous work, the use of transfer matrix method (TMM) was proposed for the simulation of the first effect. In this work, TMM is proposed to be used for the solution of Schrodinger equation with open boundary conditions to simulate the second quantum-mechanical effect. Transport properties such as transmission probability, carrier concentration, and I–V characteristics resulting from quantum transport simulation using TMM are compared with that using the traditional tight-binding model (TBM). Comparison showed that, when the same mesh size is used in both methods, TMM gives more accurate results than TBM. Copyright © 2007 John Wiley & Sons, Ltd.     

On state equations of two‐level quantum systems in a thermal environment

The state equations of open two‐level quantum systems, which form the building blocks of quantum cellular neural networks, are studied in arbitrary representations. It is shown that the dissipation matrix, that under the usual assumptions is diagonal in the energy representation, such remains if and only if the coupling between the states induced by an external field is real and infinitesimal. Copyright © 2004 John Wiley & Sons, Ltd.     

改进量子遗传算法在PID参数整定中应用

参数整定是PID控制器设计的关键,针对PID控制器参数整定问题,提出一种基于改进量子遗传算法的参数整定方法.该算法在基本量子遗传算法的基础上引入了量子交叉、量子变异和群体灾变操作.基于改进量子遗传算法的PID参数整定方法将PID控制器参数整定转化为参数优化问题,通过改进量子遗传算法的进化计算实现参数整定.与其他参数整定优化算法的仿真结果比较表明,该方法能获取更好的控制品质.仿真结果验证了该方法的可行性.     

基于新特征提取法和量子神经网络的手写数字识别

研究了一种将新特征提取方法（13维特征提取法）与量子神经网络相结合,来实现手写数字识别的方法。13维特征提取法是从每个字符中提取关键的13个特征值,而量子神经网络是将神经元与模糊理论相结合的模糊神经系统,能很好地减少模式识别的不确定度,提高模式识别的准确性。通过使用MNIST样本库仿真比较实验可知,该方法具有设计算法相对比较简单,且识别正确率较高的特点。     

基于量子神经网络的电力电子电路故障诊断

针对电力电子电路故障诊断时故障模式间存在交叉数据的模式识别问题,在量子计算和人工神经网络结合的基础上,提出了一种基于量子神经网络的故障诊断方法,并以双桥12相脉波整流电路为例进行故障诊断.实验结果表明:量子神经网络有一种固有的模糊性,它能将不确定性数据合理地分配到各故障模式中,从而使网络具有高性能、更好的鲁棒性和省时的特点,且能正确地识别大部分的样本故障模式,成功地完成电力电子电路的故障诊断.     

基于量子编码的离散粒子群算法配电网重构

阐述了配电网重构数学模型、二进制粒子群算法、量子编码的基本理论,对量子粒子群算法配电网重构进行了研究,将量子编码应用到离散粒子群算法中,用量子比特概率表示离散粒子的状态,根据二进制粒子群速度更新公式更新粒子的状态,改变开关开合状态进行网络重构.量子比特概率能够表征丰富的信息量,保证粒子的多样性和全局搜索能力.通过2个算...     

基于二进制量子粒子群算法的含分布式电源配电网重构

配电网重构是一个复杂的非线性组合优化问题。为了克服基本优化算法易陷入局部最优解的问题,提出了一种改进的二进制量子粒子群算法(BQPSO),对含分布式电源(DG)的配电网重构模型进行求解。通过引入遗传算法的交叉操作和变异操作来避免早熟来提高算法的全局搜索能力,改进了算法的性能。并且选择了适当的不可行解处理方式来提高了算法的计算效率。最后通过对IEEE33节点配电系统进行仿真,验证所提算法在求解重构问题时得到的解更好,收敛速度和全局寻优能力都有提升。     

HHT结合QPSO-GRNN的同塔双回输电线路故障测距

为提高同塔双回输电线路故障测距的精度,提出将希尔伯特黄变换和量子粒子群优化的广义回归神经网络相结合的方法用于构建测距模型.首先将线路两端采集的故障电流进行相模变换,选取其特征模量进行希尔伯特黄变换;将变换得到的2个采样点作为模型输入,对应的故障距离作为模型输出,构建经量子粒子群算法优化的广义回归神经网络;在网络中进行训...     

A novel particle swarm optimization method based on quantum mechanics computation for thermal economic load dispatch problem

This paper presents an efficient strategy to solve the thermal economic load dispatch (ELD) problem by considering several aspects of ELD. ELD performs an important role in the economical operation of power system, which essentially involves nonlinearity according to the characteristics of the generators. The complexity is amplified when the generators' prohibited zones and valve‐point effect are considered, which makes ELD a nonconvex and nonsmooth problem. The strategy employs a mechanism involving a quantum mechanics‐inspired particle swarm optimization (QMPSO). The conventional PSO is modified by integrating quantum mechanical theory which redefines the particles' positions and velocities in a dynamic manner and therefore explores more search space. The QMPSO employs a multipopulation‐based scheme which ensures particle movement and avoids premature convergence at the same time. Moreover, in order to diversify the particles, a dynamic mutation operator is introduced in the proposed method. Such features deliver a fine balance between the local and global searching abilities. Simulations are carried out by considering several cases of thermal units of varying combinations of system configurations such as with and without the valve point, with and without network loss, and for one or several hours of load demand. The results are quite promising and effective compared with several benchmark methods. © 2012 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.     

Multidimensional binary search calculation of the neuron model

Artificial neural networks have the potential for parallel processing with integrated circuit technology. Over 1 million gates are now available in the latest FPGA (Field Programmable Gate Array). However, the sum‐of‐product circuit used for evaluating inputs of the neuron model is complex and not effective for hardware implementation by FPGAs. In this paper, an improved calculation algorithm of the perceptron‐type neuron model is proposed, based on multidimensional binary search. Since the search does not need the sum‐of‐product circuit, the designed neuron circuit is small and fast. It is suitable for hardware implementation. © 2001 Scripta Technica, Electr Eng Jpn, 138(2): 24–32, 2002     

基于深度强化学习的多端背靠背柔性直流系统直流电压控制

为了提高互联配电网多端背靠背柔性直流系统的直流电压控制精度,增强抗干扰能力,提出一种基于深度强化学习的直流电压控制方法,将深度学习神经网络与确定策略梯度融合,实现连续动作搜索,自适应调整电压控制策略.首先,建立多端背靠背柔性直流系统数学模型,分析直流电压控制的非线性和不确定性特征;然后,给出了基于深度强化学习的直流电压控制算法框架,设计了动作与状态空间、奖励函数、神经网络和学习流程;最后,通过仿真分析发现,相比传统比例-积分(PI)控制方法,所提方法具有更好的动静态性能,有效提高了直流电压的控制精度,减小了扰动下直流电压波动和功率超调,缩短了直流电压和功率的恢复稳定时间.     

基于量子粒子群混合算法的电力系统无功优化

针对传统粒子群算法在无功优化中易陷入局部最优和后期收敛慢等问题,提出了基于量子粒子群混合算法的无功优化计算方法.该算法将量子叠加态思想引入到粒子群算法中,使得单个粒子能表示更多的状态和量级,增加了种群的多样性;采用量子旋转门更新粒子的速度和位置,提高了收敛的速度.用该算法对IEEE 30节点系统进行无功优化计算,并与粒...     

基于RQGA和非支配排序的多目标混沌量子遗传算法

为了提高多目标优化算法的收敛性、分布性和减少算法的计算代价,借鉴实数编码遗传算法和多目标优化理论,构建一种多目标混沌量子遗传算法.在分析量子位概率的混沌特性、量子态干涉特性和量子位实数编码的基础上,采用量子位概率交叉和混沌变异的方式进化种群,以提高寻优能力和收敛速度,利用非支配排序、精英保留和分层聚类等多目标优化策略保持种群多样性的同时,保证进化向Pareto全局最优解集方向进行.通过混合算法性能对比测试验证了多算法集成的有效性,并分析关键参数对算法性能的影响.电力系统多目标无功优化的仿真结果验证了该算法的有效性和可行性.     

基于遗传算法-BP神经网络优化的PID控制

利用遗传算法全局随机搜索能力,设计一种基于遗传算法的神经网络学习算法。对于非线性复杂系统,常规PID控制器不能获得理想的控制效果,针对复杂非线性对象的神经网络PID控制不失为一种有效的控制策略。该文提出了基于遗传算法优化参数的神经网络PID控制器,实现了基于实数编码的GA参数优化。仿真结果证明了该算法的有效性。     

基于交叉算法的神经网络模拟电路特性分析

本文利用基于交叉算法的神经网络训练方法对模拟电路进行性能分析.前馈神经网络的监督学习通常是一种从上到下(top-down)的学习模式,具有单隐层结构的前馈神经网络也可采用从下到上(bottom-up)学习模式的非监督学习算法来进行,基于交叉算法的复值神经网络训练方法突破以往算法的各种局限,其学习过程将从下到上的非监督学习和从上到下的监督学习相结合,网络性能更优.模拟电路特性分析的仿真研究表明该算法行之有效.