频率相关转换矩阵的一种平滑算法

使用传统算法得到的转换矩阵中的元素存在突变点,会给有理函数近似带来很大困难。为此提出了一种新算法能够使转换矩阵成为频率的光滑函数;新算法先使用传统的算法计算出每个频率点的转换矩阵,然后再通过一个平滑算法消除转换矩阵中由传统算法产生的突变点。在实现新算法的基础上,通过一个双回三相不换位输电线路的算例将新算法与Newton-Raphson算法进行了比较。实验结果和理论分析表明新算法具有更高的效率和更好的追踪能力。     

宽带相干源波达方向估计的新方法及性能分析

本文首先介绍了宽带阵列信号的模型,然后给出了一种新的基于信号子空间的宽带信号DOA估计算法。新算法直接分解接收数据协方差矩阵得到聚焦矩阵,通过聚焦变换将不同频率点的信号子空间聚焦到参考频率点,聚焦后得到单一频率点的数据协方差矩阵,再应用窄带MUSIC算法进行DOA估计。由于是直接分解接收数据协方差矩阵得到聚焦矩阵,因此新算法无需预估计波达方向,可以避免角度预估计对测向性能的影响,而且减少了计算量,提高了宽带信号DOA估计的实时性。最后在16元均匀线阵的情况下,选用宽带高斯信号源进行仿真,将新算法与RSS算法和TLS算法相比较,验证了新算法的优良性能。     

基于特征提取矩阵的稀疏系数求解算法

在压缩感知算法的基础上,提出了在字典学习算法过程中同时训练得到一个投影矩阵,通过该矩阵可以直接运算求取稀疏系数的方法.字典训练过程采用的是KSVD字典学习算法,并与传统的L1范数求解算法进行比较,通过实验可知,该方法比传统利用贪婪法等L1算法具有更加快速、识别率更高的特点,提出的算法通过矩阵运算可以直接求解出系数项,而后者则是一个NP问题,需要利用迭代算法来求解,这样对于大样本的测试来说提出的算法具有更好的应用空间,节约的时间将非常显著.     

近场声源三维定位MUSIC算法研究

信号到达方向(DOA)估计是阵列信号处理的最主要目的之一,基于语音阵列的DOA估计与传统的窄带远场信号的DOA估计在信号模型和处理方法上有很大区别.本文把经典的MUSIC算法推广到语音阵列处理中,实现了声源三维定位.该算法的基本思想是:把宽带语音信号通过短时傅立叶变换转换为独立离散频率点,利用MUSIC算法求出每个/部分频率点的空间谱矩阵,然后加权平均这些空间谱矩阵.通过搜索该平均空间谱矩阵,找到信号源的位置.最后,在16元均匀圆环阵的情况下,选用双语音信号源进行了仿真.仿真结果表明,该算法定位性能优良.     

宽带聚焦波束形成算法优化

通过聚焦的方法实现宽带信号的波束形成,可以得到更好的波束形成效果,而且运算量较小,精度更高。在聚焦过程中,聚焦矩阵和聚焦频率选择的偏差会直接影响宽带聚焦波束形成效果,严重时聚焦算法还可能失效,达不到宽带聚焦的目的。利用稳健Capon波束形成算法(RCB)算法对聚焦矩阵进行修正,使宽带信号通过修正的聚焦矩阵全部聚焦在最佳聚焦频率上,最后再对窄带数据利用二阶锥方法实现波束形成。仿真结果表明该算法得到的宽带聚焦波束旁瓣级更低、主瓣更窄。     

基于稳定图和矩阵束算法在风机系统低频振荡参数辨识的应用

为了精确检测含有风机的电力系统低频振荡参数,提出了一种稳定图和矩阵束相结合的新算法。首先用稳定图法确定区间低频振荡信号的阶数,在得到了系统阶数之后,就可以应用矩阵束算法得到每个分量的幅值和特征根,进而可以获得系统的阻尼比、幅值和频率。最后的仿真实验结果表明,与TLS-ESPRIT算法相比,该方法计算的阻尼比最大相对误差更小。在不同的系统中,运用该方法依然可以得到较准确的结果。     

基于预测算法和变结构的矩阵变换器驱动感应电机无差拍直接转矩控制

提出一种新型的基于预测算法和变结构的矩阵变换器驱动感应电机无差拍直接转矩控制方法,变结构控制器产生负载角度变化值,并与定子磁链给定值送入基于预测算法的无差拍直接转矩控制器中,得到预测定子控制电压矢量,结合矩阵变换器双空间矢量调制方法,实现矩阵变换器开关频率恒定,并有效地减小了转矩脉动。实验结果表明,通过使用该方法,既可以保留传统矩阵变换器驱动感应电机直接转矩控制转矩快速响应的特点,又可以保证开关频率恒定和减小转矩脉动,有效地改善了系统的动静态性能。     

联合三相锁相环和优化三点法的供电频率高精度测量算法

针对客运专线电能质量检测装置中频率测量算法精度低、计算量大和精度易受谐波成分影响等不足,提出一种以三相锁相环和三点法为核心的高精度频率算法。该方法利用三相锁相环提取频率信号,可有效克服频漂,减少所提取频率信号与理论基波信号间的偏差。详细推导了三点法求频率所带来误差的理论原因,进而得到误差抑制程度最佳时的采样点间隔,最终通过最小二乘法拟合得到频率值,进一步抑制了算法检测的误差。经多次实验证明,联合三相锁相环和优化三点法求取频率的方法,其精度和效率均优于传统频率检测算法。     

电网不平衡下基于自适应观测器的锁相环研究

电网不平衡条件下,传统软件锁相环已经不能准确跟踪电网电压的频率和相位信息,且无法直接得到电网电压的正负序分量。针对传统软件锁相环的缺陷,提出了一种静止坐标系下基于自适应观测器的软件锁相环算法。算法设计了电网电压角频率的自适应观测器模型,通过定义Lyapunov函数,推导出电网角频率的自适应律,并设计了使得观测器稳定的增益矩阵。该算法不仅可以观测到电网的频率、相位等信息,同时可以得到电网电压正负序分量。仿真和实验结果表明,基于自适应观测器的锁相环具有检测精度高、动态效果好等优点,当电网处于各种不平衡条件时更能显示其优越性,可为风电网侧变流器的算法改进提供有效的控制基准。     

改进的Prony算法在多正弦信号频率估计中的应用研究

针对噪声背景下经典Prony算法对多正弦信号频率估计受限的问题,提出一种抗噪性强的改进Prony算法。通过建立一种新的累积关系,提出的算法可以在低信噪比情况下,仅利用适当数量的新序列值线性重构稀疏和,从而较为精确地估计出信号的频率。对提出的算法与经典Prony算法的性能作出仿真实验对比。实验结果表明,在信号中嵌入噪声时,经典Prony算法在估计信号频率时失去效用,而新算法依旧可以有效估计信号频率。所以,提出的算法抗噪能力较强,性能相对稳定,精度较高,在多正弦信号频率估计中表现出更好的实用性。     

A Method for the Calculation of Frequency-Dependent Transmission Line Transformation Matrices

A stable and efficient method for evaluating the transformation matrices of both overhead transmission lines and underground cables as smooth functions of frequency is presented. The proposed method is based on the postprocessing of the eigenvectors obtained by the conventional methods where optimization is employed to maximize the smoothness of the transformation matrices. Even in the case of the underground cable systems composed of several identical single core coaxial cables, where some of the eigenvalues become arbitrarily close at high frequencies, it is capable of producing smooth transformation matrices that can be accurately fitted with rational functions.     

具有变量范围约束的潮流算法在OPF中的应用

提出一类新的电力系统最优潮流(OPF)算法.运用函数变换法模拟变量不等式约束,建立具有变量范围约束的潮流方程,并将目标函数与其上限值构成优化目标方程,与变量范围约束潮流方程一起组成扩展的变量范围约束潮流方程.该方程为非线性不定方程组.逐次收缩目标函数上限值,运用广义逆矩阵和牛顿-拉夫逊法求解,将OPF问题转化为一系列求解非线性不定方程组的一维优化逼近过程.介绍了以最小发电费用为目标的潮流优化问题和以最小有功网损为目标的无功优化问题的新算法,通过实例验证了算法的实用性.     

求解离散无功优化问题的精确连续化方法

提出了一种求解离散无功优化问题的新算法。该方法对离散变量进行二进制编码,从而将离散变量转化为若干0-1变量的线性组合,并将二进制变量约束转化为等价的互补约束,再利用非线性互补函数将互补约束转化为等价的非光滑方程,对其进行光滑化处理后,将原问题转化成可微的非线性规划问题,并采用非线性原对偶内点算法求解。此外,还对离散变量编码的原理以及应用原对偶内点法中遇到的求导难点进行了介绍,并通过几个不同规模试验系统的计算分析,验证该方法能有效处理离散变量,而且具有良好的收敛性和精确性。     

On the generation and optimization of equivalent ladder networks

The impedance scaling process and Norton's transformation are formulated in a uniform manner as a Cauer-Howitt transformation with diagonal matrices. For passive two-element-kind ladder networks the realizability conditions are reduced to linear inequality constraints. An algorithm is given to enumerate all vertices of the realizability polyhedrons which represent all equivalent networks with a minimum number of elements. In addition a cost function meeting practical requirements is stated and optimized, taking into account various orders of network complexity.     

Optimal ranking of measurements for state estimation by the gradient projection method

Optimal measurement ranking is the first and essential step in meter selection for the design of a reliable measurement system. This paper presents a computationally-efficient algorithm for the optimal ranking of measurements for state estimation. The algorithm maximizes the accuracy of estimates with respect to the measurement variances by performing a transformation on the problem formulation, and minimizing the resulting cost function subject to a set of linear constraints.

The proposed algorithm is based on the gradient projection method and includes some new computational features. Computational efficiency of the solution procedure is significantly improved by converting the constraints into a linear form through a transformation, using an analytical expression for the derivative of the cost function, employing the sparse inverse matrix technique and optimal ordering for the evaluation of the cost function and its derivatives, and simplifying the evaluation of the projected gradient vectors.

Performance of the algorithm is demonstrated on the IEEE 14, 30, and 57 bus systems.     

Efficient determination of Cournot equilibria in electricity markets

This paper analyzes the deregulated electricity market using game theory and investigates the properties of Cournot equilibrium points. A new algorithm for the calculation of the Cournot equilibrium of an electricity market in the absence of transmission constraints is presented. The algorithm is based on the transformation of the Cournot game into a three level decision-making scheme with economic signal exchange. The algorithm is tested on a hypothetical market in order to demonstrate the effects of market concentration and demand elasticity. Case studies indicate that the algorithm is very efficient computationally and therefore suitable for use with multi-period production scheduling models for the electricity market.     

Minimum correction method for enforcing limits and equalityconstraints in state estimation based on orthogonal transformations

A new method is presented for orthogonal transformation-based power system state estimation that enables enforcement of limits and equality constraints. The enforcing of constraints is based on minimization of the correction by which the unconstrained state estimate must be modified in order to satisfy all equality and inequality constraints. An estimation algorithm is developed which combines active set and interior point optimization methods in such a way, that positive features of both methods are retained. A self-scaling Givens rotation algorithm is proposed to perform elementary orthogonal transformations in unconstrained state estimation. This reduces the necessity of the underflow monitoring and re-scaling during the orthogonal transformation of the Jacobian matrix. The numerical results are presented demonstrating the efficiency of the algorithm     

基于矩阵束算法的谐波和间谐波参数估计

电力系统中非线性负载的不断增加导致谐波污染问题日趋严重。基于矩阵束算法提出了一种谐波和间谐波参数估计的新方法。该方法将含噪声的谐波、间谐波信号构造为两个Hankel矩阵,通过奇异值分解和矩阵的低秩近似等方法抑制了噪声的干扰,同时减少了计算量;利用矩阵间的特定关系获得频率信息后,再通过最小二乘法得到幅值和相位。仿真结果表明,该方法具有精度好、运算率高、抗噪性强等特点。     

Unit commitment with ramp multipliers

This paper presents a new decomposition method, based on the Lagrangian relaxation technique, for solving the unit commitment problem with ramp rate constraints. By introducing an additional vector of multipliers to represent the cost of “system ramping demand”, this method can handle the coupling constraints between time periods while still keeping the simplicity of the original decomposition method. A new algorithm for updating multipliers is also proposed. Similar to the bundle algorithm, this algorithm maintains the previous iteration history to approximate the dual envelope. Unlike the bundle algorithm, this new algorithm generates an update step along the subgradient direction without any quadratic programming (QP) code. The new algorithm combines the bundle algorithm's smooth approach to the dual optimum with the sub-gradient method's fast update