摘要
广域测量系统(WAMS)作为一种测量手段,不可避免地存在测量误差。为了获得更优的控制策略和分析结果,有必要对实际量测数据进行滤波处理后再应用。文中提出了一种对实际量测数据进行动态滤波估计的新方法,在发电机二阶动态方程的基础上,建立了发电机动态状态估计模型。考虑到模型的非线性,文中应用粒子滤波(PF)算法。为解决计算占用空间和计算量较大、样本退化的问题,在基本PF算法的基础上引入序列采样重要性重采样(SIR)方法。同时,还应用扩展卡尔曼滤波(EKF)算法进行了状态估计并将结果与文中所提方法进行了对比分析。为了定量评估估计效果,建立了基于估计路径相似性的评价指标。最后,通过对CEPRI 7节点系统的仿真计算,表明基于PF的估计结果与实际结果相关性较高、与真实值的均方根误差小,优于EKF的估计结果,有效减小了误差数据的影响。
As a kind of measuring method,wide area measurement system（WAMS）inevitably has measuring errors.In order to obtain a better control strategy and more accurate analysis results,it is necessary to do filtering processing for the actual measured data before application.A new method of dynamic filtering estimation for the actual measured data is proposed,and a dynamic state estimating model of a generator is established based on the second-order dynamic equation.Considering the nonlinearity of the model,the particle filtering（PF）with the unique advantage of dealing with non-linear and non-Gaussian stochastic system estimation problems is also used.To solve the problems of computation taking up space,large amounts of calculation and sample degradation,sequential importance resampling（SIR）is introduced on the basis of PF algorithm.Meanwhile,extended Kalman filter（EKF）is also adopted to do the state estimation,with the results compared with those using the approach proposed.Besides,to measure the estimation effect quantitatively,an evaluation index system based on the estimate path similarity is set.Finally,through the simulation calculation of a CEPRI 7-bus system,it is shown that the estimation based on the PF has a higher correlation with the actual result and a smaller error compared with the root mean square of the real value,a result better than the estimation based on the EKF and effectively reducing the influence of erroneous data.
出处
《电力系统自动化》
EI
CSCD
北大核心
2016年第2期49-54,共6页
Automation of Electric Power Systems
基金
国家自然科学基金资助项目(51177010
51377017)~~
关键词
机电暂态
状态估计
粒子滤波
路径相关性量度
electromechanical transient
state estimation
particle filter
path correlation measure
