Dynamic State Estimation For Smart Grid With Hybrid Measurements

With the research and construction of smart grids receiving widespread attention and concern,more and more human and material resources are invested in the construction of smart grids.The high penetration of distributed power generation has played a positive role in peak load regulation,and as a backup energy,it has also improved the flexibility of the power grid.The wide introduction of various types of intelligent measurement equipments such as Phasor Measurement Unit(PMU)and smart meters provide richer measurement information for the smart grid and is more conducive to grasping the grid operating status in real time.However,the development of smart grids also brings new challenges.The randomness of distributed power generation has brought fluctuations to the grid power flow.And how to make full use of massive phasor measurement information is a problem that needs to be researched and solved.In view of the above situation,this dissertation studies the problem of dynamic state estimation of smart grid in a hybrid measurement environment,and discusses the theoretical issues such as state estimation model establishment,estimation method selection,and model parameter optimization.Power system state estimation methods include static state estimation that has been running in actual systems for many years and dynamic state estimation that stays in the theoretical research stage.This dissertation first analyzes the static and dynamic estimation models commonly used in power system state estimation,and discusses various The different problems addressed by static estimators.Then this dissertation investigates the advantages and disadvantages of various dynamic estimation methods in terms of adaptability,calculation accuracy and calculation complexity,respectively.Secondly,the power system has long been considered to operate in a quasi-steady state,but sudden load changes occur from time to time,which is more common in smart grids.Based on this,this dissertation adopts a data-driven dynamic model and a physical-driven dynamic model of the power system.The data-driven dynamic model is more suitable for the traditional scenario where the power grid runs in a quasi-steady state and is sensitive to process noise parameters,while the physical-driven dynamic model is suitable for scenes with large load fluctuations and relatively insensitive to process noise parameters.In view of the existence of multiple hybrid measurements such as SCADA(Supervisory Control And Data Acquisition)measurements,PMU measurements,and smart meter measurements in the smart grid,multi-sensor fusion estimators are constructed for the smart transmission grid and smart distribution network,respectively.Then we make full use of high-sampling frequency sensors,establish switching rules based on the innovations of high-frequency measurement,and construct a switched system model of smart grid.In the case of different load fluctuations,the effectiveness and robustness of the proposed model to different process noises are proved.In the face of the situation that the smart meter measurements in the distribution network system are unavailable at certain state estimation moments,the influence of the load forecasting method and the multi-step prediction technology on the state estimation performance is compared and studied.When fusing measurements on different time scales,even if the PMUs are not configured in large numbers and the sampling rate of traditional measurements remains unchanged,the fusion state estimation on a small time scale is still completed.Next,on the basis of the previously established switched system model,the characteristics,adaptation range and approximate error of Extended Kalman Filter(EKF)and Unscented Kalman Filter(UKF)are contrastively analyzed.The influence of the process noise covariance of the system on the estimation accuracy of the two is investigated.In terms of the detection and identification of bad data,the chi-square test and the maximum standardized residual test method in static state estimation are reviewed.The standardized innovation test method in dynamic state estimation is used to detect the sudden load change and bad data in the system,and the skewness test is applied to distinguish the difference between the two situations.Finally,considering that process noise covariance parameters have an important influence on dynamic state estimation while they are not easy to obtain,a joint optimization method for state estimation process noise parameters based on symbiotic organism search(SOS)algorithm is proposed.The process noise covariances of the estimators based on SCADA and PMU measurements composing organisms,and taking the average root mean square of the innovations as the cost function,the process noise covariance values are jointly optimized.Aiming at the problem of slow response of SOS algorithm to the sudden load change,the SOS algorithm with variable observation window length is proposed.Experiments show that the proposed method is better than the covariance matching method in estimation and numerical stability.Finally,the influence of the observation window length on the state estimation based on covariance matching and SOS algorithm is discussed.