【作者】 张彤；

【导师】 刘建昌； 于海斌；

【作者基本信息】 东北大学 ， 控制理论与控制工程， 2021， 博士

【摘要】 随着配电系统规模的不断扩大和复杂性的日益提高,有效的故障诊断方法是保证其安全运行、改善供电质量和提高经济效益的关键。由于系统拓扑结构的相关性及状态变量的复杂性等因素,从配电系统运行过程中采集的同步相量信息具有非线性、低信噪比、动态性和相关性等特点。针对这些特性,本文对基于同步相量信息的配电系统故障诊断方法进行了深入研究,取得了系列创新性研究成果,主要工作和贡献如下:(1)配电系统同步相量信息处理方法研究。首先,基于坐标变换理论利用二阶广义积分器锁相环输出信号相互正交的性质构建同步相量信息的基波正交分量;然后,在配电系统两相正交坐标系下,基于扩展卡尔曼滤波方法提出预测-量测-校正递归框架下的准最优估计信息处理方法,将期望误差协方差迭代过程建模为黎卡提方程并采用解析方式求解得到卡尔曼增益系数,有效解决系统畸变条件下同步相量信息基波分量和高次谐波分量的跟踪估计问题;最后,在低信噪比、高次谐波畸变的IEEE 34节点系统中,通过仿真实验来验证所提方法的性能。(2)基于改进RBF神经网络的配电系统故障检测方法研究。首先,通过构造RBF神经网络线性回归模型,将隐含层中心矢量选择问题转化为显著回归矩阵选择问题,在正则化准则下利用零阶正则化误差函数约束神经网络模型有效复杂度,并利用正则化误差率构建有效隐含层中心矢量选择条件,建立具有良好泛化能力的正则化径向基(regularized radial basis function,RRBF)神经网络模型;然后,针对配电系统同步相量信息特征学习,在监督多模型残差生成分类学习框架下提出基于RRBF神经网络的故障检测方法,对配电系统故障区域和故障相进行检测;最后,在不同短路故障工况下,通过仿真实验验证所提方法能够准确检测故障区域及故障相。(3)基于幅相特性的配电系统故障特征提取方法研究。首先,在已有的网络拓扑结构上将故障点扩展为虚拟节点,利用虚拟阻抗法建立配电系统的扩展阻抗矩阵来分析配电线路拓扑相关性,提高全连通阻抗矩阵的计算效率;然后,根据金属性短路故障过程中配电线路模型的相量参数变化规律,有效融合相角信息进行分布参数模型电压电流相量特征的解析分析,提出基于幅相特性的故障特征提取方法;最后,在IEEE 34节点配电系统中进行仿真实验,验证所提方法的有效性和响应速度。(4)基于复杂度信息熵的配电系统故障定位方法研究。首先,为了分析配电线路幅相特征向量的分布特性,基于复杂度分析原理对信息熵方法进行扩展,建立复杂度信息熵,对特征向量的总体概率分布不确定性和内在复杂度进行定量分析;然后,沿配电线路分布生成定位函数来动态分析幅相特征向量的分布特性,利用复杂度信息熵的极值性建立故障定位边界条件进行故障位置定位,进而提出基于复杂度信息熵的故障定位方法;针对迭代计算步骤复杂的问题,采用Fibonacci搜索算法来改进故障定位方法的计算效率和响应速度;最后,在IEEE 34节点配电系统和IEEE 13节点电力系统末端支路中进行仿真实验,验证所提方法的定位精度和响应速度。更多还原

【Abstract】 With the increasing scale and complexity of the distribution system,the fault diagnosis method is the key to improve the safe operation,the power supply quality and the economic benefit in the distribution system.Considering the topological structure correlation and the state variable complexity,the synchronized phasor information measured in the operation process has the nonlinearity,the low signalto-noise ratio(SNR),the dynamic and the correlation.According to the data characteristic,the fault diagnosis method of the distribution system based on the synchronized phasor information are deeply studied in this dissertation.The innovative research results have been achieved and the main contributions are as follows:(1)The information processing method of the synchronized phasor is studied.Firstly,based on the coordinate transformation theory,the orthogonal fundamental component is constructed using the orthogonal property of the phase locked loop with a second order generalized integrator.In the orthogonal coordinate system,a quasi-optimal estimation information processing method is proposed based on the extended Kalman filter in the framework prediction-quantification-correction.The expected error covariance iteration process is modeled as the Riccati equation.The Kalman gain coefficient is calculated in an analytical way.The fundamental and high harmonic components are effectively estimated under the distortion condition.In the IEEE 34-bus distribution system with the low SNR and high harmonic distortion,the simulation results show that the proposed method can improve the accuracy of the synchronized phasor estimation.(2)The fault detection method of the distribution system based on the improved radial basis function(RBF)neural network is studied.Firstly,the linear regression model of the RBF neural network is constructed.The hidden layer center vector selection problem is transformed into the significant regression matrix selection.In the regularization criterion,the regularization error function is used to restrain the network effective complexity.The regularization error rate is used to construct the effective hidden layer center vector selection.The regularized radial basis function(RRBF)neural network model is constructed with a well generalization ability.Then,in the feature learning of the synchronized phasor information,the fault detection method based on the RRBF neural network is proposed in the framework of the supervised multi-model residual classification.Finally,the simulation experiments verify that the proposed method can accurately detect the fault section and fault phase under different short-circuit fault conditions.(3)The fault feature extraction method of the distribution system based on the amplitude-phase characteristic is studied.Firstly,the fault bus is extended as the virtual node in the existed network topology.The extended impedance matrix is established using the virtual impedance to analyze the topological correlation.The calculation efficiency of the fully connected impedance matrix is improved.In the metal short-circuit fault process,according to the variation law of phasor parameters,the analytical expression of the phasor characteristic is analyzed using the distribution parameter model.The fault feature extraction method based on the amplitude-phase characteristic is proposed.The accuracy and the response speed of the proposed method are verified in the IEEE 34-bus distribution system.(4)The fault location method based on the complexity entropy is studied.Firstly,the distribution characteristic of the amplitude-phase feature vector is analyzed.The entropy theory is extended based on the complexity analysis principle.The fault complexity entropy is constructed to analyze the general distribution characteristic and the intrinsic complexity of the amplitude-phase feature vector.Then,the location function is generated along the distribution line to dynamically analyze the distribution characteristic of the amplitude-phase feature.The fault location boundary condition is construted using the extremum quality of the complexity entropy.The Fibonacci search method is used to improve the calculation efficiency and response speed.Finally,the location accuracy and response speed of the proposed method are verified in the IEEE 34-bus distribution system and the bottom terminal of the IEEE 13-bus power system.更多还原