Neuro - Fuzzy Elman Network for Short - Term Electric Load Forecasting

The problem of short-term electric load forecasting (STLF) is considered. A modified architecture of Elman-type recurrent neural network is proposed. It utilizes a special fuzzification layer to deal with quantitative as well as ordinal and nominal data. The second hidden layer of the network consists of standard Rosenblatt-type neurons with sigmoidal activation functions. The context layer is formed by delay units and feeds the output signals of the second hidden layer back to its inputs. The output layer contains a single neuron with sigmoidal activation function. Several modifications of a learning algorithm for this architecture are derived based on Widrow-Hoff, Levenberg-Marquardt, and ChanFallside procedures. The proposed approach was tested on historical electric load data from several energy systems and showed promising results with respect to forecasting accuracy and speed of learning in comparison to feedforward neural and neuro-fuzzy systems. Index Terms Short-term electric load forecasting, recurrent neural networks, Elman network, learning