Research On Measurement-based Load Modeling And Simulation Validation In Large-scale Power Grids

Owing to the security in power systems,generally the researches on dynamic characteristics of power systems depend heavily on simulations.Thus,simulation reliability has large impacts on almost every aspects in power systems,such as operation,planning and design etc.Herein, validity of models and parameters in simulation guarantees the simulation reliability.As the wide applications of WAMS(Wide Area Measurements System),it is possible to evaluate simulation veracity as well as validate and calibrate models and parameters making use of the measured data. However,there are still many difficulties impeding the work of simulation validation in large-scale power grids,which focus on high-dimensionality,strong non-linearity,complexity of network structure,immense electric components and so on.The artificial major disturbance tests in NE power grid of China provide valuable measured data regarding the load dynamics and system dynamics.Thus,combined with our experience on load modeling,firstly,this thesis proposes the principles and methods on measurement-based load modeling at the first time.The cases study validates the measurement-based load modeling and demonstrates its good generalization capability both in substation sense and power grid sense. Then,according to the time-domain and frequency-domain characteristics of power systems,a set of error criteria is proposed to quantify simulation discrepancies and to evaluate simulation veracity based on similarity theory,analytical hierarchy process and other methods.When there are discrepancies between simulations and measured data,based on hybrid dynamic simulation via the Variable Impedance Method,the simulation validation strategy is presented to decouple a big power grid into several small areas with fewer components. Therefore,the simulation validation work can be carried out in a much specified area and the wrong component raising the error can be found.Since one type of electric component has many models with large number of parameters and the sensitivities of measured data respect to each parameter are different,after locating the erromeous component,the trajectory sensitivity method is applied to analyze the parameter sensitivity of the composite load model and generator model. Also the parameter identifiability and calibration based on trajectory sensitivity are discussed.Actually,there are lots of uncertainties in real systems,such as model approximation, algorithm error etc,which have great impact on dynamic stabilities.Thus,a detailed analysis of model and parameter uncertainty will identify the key sources of uncertainty that merit further research,which could provide useful information and guidance for the modification and calibration work.However,how to quantitatively analyze the uncertainty arising from model structure and parameters is still hard to realize,mainly because the traditional methods are inapplicable to power system dynamic simulations for their repeated simulations and intensive computation time.Therefore,this thesis introduces two new methods to quantitatively analyze the uncertainty of load models and its influence on simulation outputs,namely,the Probabilistic Collocation Method(PCM) and the Stochastic Response Surface Method(SRSM).Whereafter, according to the uncertainty analysis,the thought of locating key parameters and key areas that heavily affect the dynamic stability is presented,in order to reduce the space for simulation validation.This thesis focuses the researches on increasing simulation veracity and enhancing model and parameter validity,which the final purpose is to improve simulation reliability.Herein,relative solutions and methods suitable for large-scale power grids are proposed,following the stages of modeling-error evaluation-error location-model and parameter calibration-uncertainty analysis. Cases study of simulation system and real system validates the proposed solutions and methods.