Generalized Power Load Modeling Method with Multiple RBF Neural Network Model Structures

This paper proposes a new generalized power load modeling method, which contains multiple RBF neural network model structures. Firstly, based on the REF neural network, a sub-model is built to describe typical load characteristics. It breaks through the limitations of load components, and adapts to the diverse development of modern power system. Then, based on Bayesian estimation theory, multiple typical load models arc merged into one integrated load model. It solves the problem of insufficient generalization ability of traditional load model, and adapts to the time variation of power load. Finally, the proposed method was applied in IEEE 14-bus test system. The results obtained proved its validity.