Theoretical Study Of Power System Load Modeling And Application Of Ann Load Model

It is the consensus of power circle that power system load modeling is a very difficult task, and has drawn many attentions of researchers from home and abroad. Load modeling has a significant impact on power system digital simulation, as well as on the layout, design, operation and control of a power system. So far two modelling approaches have been developed, namely so-called component-based and measurement-based ones. From the aspects of modelling structures, modelling by mechanism and modelling by input/output are formed respectively. Artificial neural network (ANN) was implied into load modeling for its excellent nonlinear approximation capability. This paper emulates dynamic load behaviors through a time-domain simulation on induction-motors, and utilizes ANN to approximate their characteristics. Two examples are represented to illustrate the proposed viewpoints.The core of this paper is listed as follows:1) Combining the advantages of both component-based and measurement-based methods, this paper makes a cluster and synthesis of load models. Guidelines for load clustering are developed, and the theory of pattern recognition is introduced to load clustering.2) At present, the reactive accuracy of most load models is rather low comparing to the active accuracy. The reactive and active power of a load group should be considered simultaneously rather than respectively. Combining the advantages of both modeling by mechanism and by input/output, this paper proposes a power coupled load model in order to increase the reactive accuracy of the model.Simulation results justify the cluster and synthesis approaches of load models, and the proposed power coupled load model could be valid in increasing the reactive accuracy of a load model.