The study on composite load model structure of artificial neural network

BP artificial neural network is suitable for any nonlinear curve fitting at any precision theoretically and occupies simple structure. So it is widely used as the model of power composite load in past load modeling research. In this paper, the authors point out that BP ANN is not suitable for describing the dynamic characteristics of power composite load because of two reasons. One reason is due to the network structure without feedback, the other reason is due to the training algorithm of BP. The feature and disadvantages of BP ANN load model is discussed in detail in the paper. Aiming at the special requirement of composite load modeling, the authors propose an ANN composite load model based on Elman artificial neural network. The authors illuminate the structure feature, mathematic description, identification strategy and modeling method of this model. Then we make modeling tests with a number of measured data at substation field. The results obtained by modeling tests show that the composite load model based on Elman ANN occupies a series of advantages such as simple structure, less parameters, convenient in engineering application, strong ability describing composite load and so on. This paper is divided into five sections. Section 1 is introduction which gives a simple overview of ANN load modeling and states importance of ANN load modeling. The feature and disadvantages of BP ANN load model is systematically discussed in section 2. Section 3 illuminates the structure feature, mathematic description, identification strategy and modeling method of Elman ANN load model. Some examples of composite load modeling with BP ANN and Elman ANN load model respectively in section 4. Section 5 is the conclusion. Section 6 is a list of recent references.