Research On Power System Dynamic Load Modeling And Model Validation

This paper studies some important problems in three major respects of Measurement-Based Load Identification Method: Time dependence characteristic of load, mechanism of synthetic dynamics and model validation. This gives significant contribution to the application of the load models. The main points of the research work of this thesis are as follow: Becauce of load time dependence characteristic, different models, whch have different order and parameters, can be built for different set of field data, this results in is a difficulty in choose parameters. So a two-step concept in load modeling is proposed as follow: The first step is to identify the most likely model structure to describe different sets of measured data, instead of pursuing a perfect fitting result from one set of data. The second step is to analyze the distribution of parameters of load model, which share the same model structure, by statistical method. It was difficult to describe different reactive load characteristics by linear discrete models of same order in the past. Therefore the elements that might affect the structure stability of the linear discrete model is studied. An effective solution to order instability of the reactive load model is proposed for the first time. Base on the above research, the author studied the possibility of utilizing the same model structure to describe different set of field data. The average fitting level against different measured data between the 1 order and 2 order difference equation models has been compared. The study reveals that 1 order equation can describes load characteristics well, both on real power and reactive power. The parameters of difference load model are not suitable to be considered as character vector because meanings of them is not intuitionistic. and varys along with the sample cycle, therefore this thesis presents the methodology for representing load characteristics with character vector. The character vector is composed of three characteristics of load step response. With statistical theory and methodology, the probability distribution functions of load characteristics parameters are analyzed. The analysis of the parameter probability distribution is significant for selecting reasonable parameters to utilize the load model As for the same load spot, there are many different load characteristics; it is usually to use the most typical load characteristic in practice, so the author study two methods to identified the general load characteristics for the samne load spot. The description ability to different data using general load characteristic , which is obtained by two methods, are compared. Based on two real examples from Guangdong Power Grid, the author studied the validation of the non-mechanism load dynamic model. The result shows that the Guangdong power load dynamic model is affective to both stability limit calculation of the power across exporting from Yuexi and simulation of faults in Shuibei substation. The former came to a very close result between experienced static model and the dynamic model from field data. For the later, the dynamic model is closer to actual recorded wave than static model. The study confirmed that different parameters for dynamic models would great impact on the calculation result and dynamic model would greatly increase calculation accuracy. The induction motor dynamic load model (TM DLM)is supported by most power system simuation program, but the model structure of current TM_DLM is not suitable to describe synthetic load, therefore it is...