Load Modeling Of Large-scale Power Grid Considering Time-variant Characteristic

Load modeling is one of the most difficult problems in power system analysis. There have been lacking the systematic methods with wide applicability in this field. This dissertation proposes, in the first time, three progressive layers to model the dynamic load, namely, modeling on load basic characteristics, modeling on load time-variant characteristics and modeling on the load of large-scale power grid. The first layer focuses on the modeling on the dynamic characteristics of the specific physic load at the specific time; the second layer focuses on the modeling of the general dynamic characteristics of the specific load all the time; while the third layer concentrates on the load modeling for large-scale power grid.The key in modeling the load basic characteristic lies on the choice of the model structure, the identification algorithm and also on the stability of the model parameters. After widely researching on the currently available typical load models, this dissertation presents the Time-variant Adaptive load model structure, i. e. TVA load model structure, which is more adaptive to the measurement-based load modeling. Based on the presented model, the dissertation further researches and applies the Least Square Algorithm in classic system identification theory as well as the Simplex Method and the Genetic Algorithm in modern system identification theory to identify the load model. Furthermore, the dissertation researches the identifiability of the load by analyzing qualitatively the non-unique characteristic of the load parameter, and gives the numeric zone of the load parameters. The dissertation also investigates the view of the load model from different voltage levels.The most difficult aspect of load modeling, which differentiates itself from other modeling problems, lies in the fact that the load is time-variant. The dissertation proposes originally the idea of decomposing the time-variant load to solve this problem. First, the time-variant load is decomposed into the time-variant load characteristic and the time-variant load model due to the modeling method. Second, the multi-curve approximation method is proposed to identify the load parameters, which to a great extent, reduces the effects on the time-variation of the load model from modeling method. Then, time-variant load characteristic is decomposed and the TVA load model structure is proposedto reduce the effects on the time-variation of the amplitudes of the load. Finally, the group algorithm and Hypothesis Testing strategies are applied to solve the problem of the time-variation of the load components. The whole problem related to the time-variation of the load is completely solved by the proposed method, the core of which lies in the TVA model structure.From the view of the engineering application, large-scale power grid load modeling addresses modeling the load distributed at various sites. After analyzing the inefficiency of the group algorithm for modeling load for large-scale power grid, the dissertation proposes a load modeling approach for large-scale power grid based on the aggregation theory, thus solves the measurement-based large-scale power grid load modeling.The theoretical results proposed in this dissertation have been applied in engineering practices. An advanced software platform with the parameter identification as its core has been developed for load modeling. This software platform has been practiced in the engineering project of load modeling for Zhangjiakou area, as well as the National Power Grid key project of load modeling for the North East of China. Furthermore, combining with the survey on the components of the load in the North East of China, a more accurate load model database of the North East of China is being built.