Short term load forecasting using fuzzy neural network modified by the similarity and subsethood measures

In this paper the linguistic approximate reasoning method is used for short term load forecasting. A neural structure for inference processing units is put forward. Two different Analogical and Deductive - approaches to the inference method have been distinguished. Correspondingly, two different architectures - Analogical and Deductive fuzzy neural networks - are introduced in this paper. The accuracy of the load forecasting, as well as the size of the required rule base have been compared for Analogical, Deductive, and conventional fuzzy neural networks. It is shown that Analogical and Deductive approaches have superior performance in this application.