Build the regional power grid load model dynamic database using ANN and multi-agent

This paper presents a novel approach to build the regional power grid load model dynamic database (LMDD) using artificial neural network (ANN) and multi-agent technique. As the component-based approach and measurement-based approach have their own demerits, dynamic load curve analysis method which is based on Kohonen clustering and load characteristics is proposed to gain the percentages of load classes on feeders or transformers. Through this method, it is possible to overcome the time varying of load characteristics. To check the validity of created load model, an elementary verifying means has been discussed. In the process of build LMDD, block structure and multi-agent technique are used to ensure the extension of the system function easily in future. The establishment of LMDD will offer reliable load model for power system calculation and simulation, and it will benefit the power system to be in the safe and stable operation. (6 pages)