Prediction of Chaotic Dynamics Parameter for the Flow of Overhead Line Failures

Abstract

The results from predicting the failure rate of 500 kV overhead power lines represented as a time series and having signs of its being chaotic in nature are presented. The predictive estimates were obtained using the regression (spectral singular analysis) method and artificial intelligence (neural and fuzzy neural networks) method. The matrix of delays formed proceeding from the failure rate time series was used as the object of spectral singular analysis. The prediction was made by carrying out single-step transformations of input data. For carrying out prediction by means of a neural network, a direct signal transmission network trained using the back propagation method was used. For achieving the minimal root-mean-square error, the training sample contained the maximum possible prehistory. For predicting the failure rate using the fuzzy neural network method, the Wang—Mendel neural network was chosen. It is shown that the predictive estimates of the failure rate obtained with a sufficiently large number of experiments on the basis of a «highly developed» neural network lead in fact to fulfilling the conditions of the central limit theorem. As a result, the failure rate time series prediction methods based on using regression and artificial intelligence yielded different estimates. This outcome testifies that this is only the beginning of work on substantiating the methodology for predicting failure rates in electric networks.