Use of artificial neural network in prediction of electrical load

Бесплатный доступ

Every year there is an increase in development followed by an increase in the demand for electricity. Because of this, it is necessary to forecast the number of consumers and the demand for electricity so that electricity providers can provide electricity as needed and minimize outages. The forecasting uses an artificial neural network (ANN) backprop method. The advantage of this method is the convenience in formulating predictor data, as well as the flexibility to change. The Levenberg-Marquardt learning algorithm, a variable Greendent and quasi-Newton learning rate are used. The most accurate results can be seen by looking at the lowest average error rate obtained with all three learning algorithms. The results of training with trainlm, traindx, and trainbfg show that the resulting error rates are quite small at 0.194%, 0.15%, and 0.14%. The training results show that the artificial neural network (ANN) is well suited for predictive applications.

Еще

Artificial neural networks, consumption, electricity, learning algorithm, prediction, back propagation method, denormalization

Короткий адрес: https://sciup.org/147241064

IDR: 147241064

Статья научная