Neural Network Prediction of the Electricity Consumption of Trolleybus and Tram Transport in Sofia City

摘要

A multilayer neural network with back propagation learning method used for prediction of the electricity consumption of the trolleybus and tram transport in Sofia city is presented in the paper. In the hauler "Transenergo" the Power Engineer has to declare necessary electricity consumption for every hour of the following week. The incorrect request affects the price of the electricity. Electricity consumption is a random process which depends on many factors. The Power Engineer has information for the following data: kilometers run, temperature, the kind of day and from this information has to declare the necessary electricity consumption. In this paper the property of neural networks to define the relation among a number of variables is employed for solving the problem of electricity consumptions prediction. The data for training of the NN are for the 2011-2012 period (17493 items). Testing data is for the 2013 year. The neural network has one input with five neurons, one hidden and one output layer with one neuron. The output is the electricity consumption. The accuracy, which is reached, is bigger than the Power Engineer has achieved in the real request for electric consumption.