电力系统分布式动态状态估计研究

发布时间：2018-01-12 17:04

本文关键词：电力系统分布式动态状态估计研究　出处：《华北电力大学》2014年博士论文　论文类型：学位论文

【摘要】：同步相量测量单元(phasor measurement unit, PMU)能够对电力系统机电暂态过程中相量信息进行直接测量,为电力系统动态安全监控提供了新的技术手段。然而,由于传感器误差以及干扰的影响,PMU量测不可避免地存在随机误差和不良数据。如果不对PMU量测量进行处理而直接应用,则有可能无法准确监测电力系统动态过程,甚至导致控制系统做出错误的控制策略。针对PMU量测信息,本文系统地研究了机电暂态过程中分布式动态状态估计方法。论文主要研究成果如下： (1)提出分布式动态状态估计框架。在发电厂和变电站分别进行发电机动态状态估计和变电站零阻抗特性状态估计,将估计结果上送至调度中心进行数据整合并实施全系统状态估计。提出了一种系统机电暂态过程中基于PMU的发电机动态状态估计新方法。该方法充分考虑系统机电暂态过程中调速器对发电机机械转矩的调节作用。建立了系统机电暂态过程中发电机动态状态估计模型；给出了系统噪声误差方差的具体计算方法；进一步提出基于比例对称采样无迹卡尔曼滤波的发电机动态状态估计算法。仿真结果表明提出的方法精度高于机械转矩恒定的方法。 (2)针对PMU量测中存在不良数据的问题,提出一种鲁棒性发电机动态状态估计算法。将时变多维观测噪声尺度因子引入到容积卡尔曼滤波中,根据量测新息对量测误差进行在线调整,使其更加逼近真实噪声。再利用调整后的误差计算滤波增益,使其能够在PMU量测存在不良数据的情况下对状态量预报值进行准确修正从而得到精确的发电机状态量估计值。针对时变多维观测噪声尺度因子为非对角阵而造成滤波增益求逆发生奇异的问题,提出解决方案。仿真结果表明,当PMU出现连续多点坏数据时,鲁棒动态状态估计仍然能够得到准确的估计结果。 (3)提出了一种系统机电暂态过程中,基于PMU的变电站状态估计新方法。该方法将变电站内断路器的零阻抗特性作为虚拟量测,进一步提升冗余度。同时,在系统故障后断路器状态未知的情况下建立状态估计模型,能够有效辨识断路器的实际状态。针对PMU量测存在不良数据的问题,给出了基于非二次准则状态估计的不良数据辨识方法,并对门槛值的选取方案进行了改进,能够有效辨识不良数据。 (4)提出了一种机电暂态过程中全系统状态估计方法。基于机网接口的直接解法,给出了发电机动态状态估计结果转化为网络节点电压相量伪量测的误差方差计算方法；提出了考虑发电机动态状态估计约束的全系统状态估计方法,通过发电机动态状态估计约束进一步提升机电暂态过程中系统状态量的估计精度。
[Abstract]:The synchronous phasor measurement unit (PMU) can directly measure the phasor information in the electromechanical transient process of power system. It provides a new technique for power system dynamic security monitoring. However, due to sensor error and interference. PMU measurement inevitably has random errors and bad data. If the PMU measurement is not processed directly, it may not be able to accurately monitor the dynamic process of power system. This paper studies the distributed dynamic state estimation method in electromechanical transient process systematically. The main research results are as follows: 1) A distributed dynamic state estimation framework is proposed. Generator dynamic state estimation and substation zero-impedance characteristic state estimation are carried out in power plants and substations respectively. The estimation results are sent to the dispatching center for data integration and the whole system state estimation is implemented. A new method of generator dynamic state estimation based on PMU in the electromechanical transient process of the system is proposed. The method takes full account of the system. The dynamic state estimation model of generator in electromechanical transient process is established. The calculation method of system noise error variance is given. Furthermore, a generator dynamic state estimation algorithm based on proportional symmetric sampling unscented Kalman filter is proposed. The simulation results show that the proposed method is more accurate than the method with constant mechanical torque. 2) aiming at the problem of bad data in PMU measurement, a robust dynamic state estimation algorithm for generator is proposed. The time-varying multidimensional noise scale factor is introduced into the volumetric Kalman filter. The measurement error is adjusted online according to the measurement innovation to make it more approximate to the real noise. Then the filter gain is calculated by using the adjusted error. It can correct the prediction value of state quantity under the condition of bad data in PMU measurement, and get the accurate estimation value of generator state quantity. The scale factor of noise is non-diagonal matrix for time-varying multi-dimensional observation. The singular problem is caused by the inverse of the filter gain. The simulation results show that the robust dynamic state estimation can still get accurate results when the PMU has continuous multi-point bad data. In this paper, a new method of substation state estimation based on PMU is proposed, in which the zero impedance characteristics of circuit breakers in substation are taken as virtual measurements. At the same time, when the state of circuit breaker is unknown after the system fault, the state estimation model can effectively identify the actual state of the circuit breaker. The problem of bad data exists in the PMU measurement. In this paper, an identification method of bad data based on non-quadratic criterion state estimation is presented, and the selection scheme of threshold is improved, which can effectively identify bad data. A state estimation method for the whole system in the electromechanical transient process is proposed, which is based on the direct solution of the machine network interface. An error variance calculation method is presented in which the dynamic state estimation results of the generator are transformed into the pseudo-measurement of voltage phasor of the network node. A full system state estimation method considering the constraints of generator dynamic state estimation is proposed. By using generator dynamic state estimation constraints, the estimation accuracy of system state variables in electromechanical transient process is further improved.
【学位授予单位】：华北电力大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TM31

【参考文献】