NEURAL NETWORK MODEL FOR LOAD SHAPE FORECASTING IN DISTRIBUTION ELECTRIC SYSTEMS

摘要

A neural network (NN) based model is proposed in this paper for short-term load shape forecasting in distribution electric systems. At the distribution level the neural model has to cope with low load aggregation, which means poor defined patterns and random behavior of the loads. The modeling is performed in two steps. The first one uses a Kohonen NN for type of day classification and bad data detection. In the second one the load shape is forecasted by a multilayer perceptron NN. The model has been validated using real data and is specially well suited for the applications emerging from the new structure of the power sector. Temperature influence on the load, and model comparisons has also been evaluated in this work.