Short-term Load Prediction Based on Chaos Time Series Theory

In this paper, two chaotic predicted methods are applied to forecast the grid's load data. The data are collected from the grid of New South Wales, Australia. It records the grid's load of four weekends in May. First, the phase space is reconstructed using the delay embedding theorem suggested by TAKENS. Second, for reducing the negative influence of the Largest Lyapunov Exponent Method, a method based on the Adding-weighted Largest Lyapunov Exponent Method is proposed. Then the Adding-weighted One-rank Local-region Forecasting Method as a traditional chaotic forecasting arithmetic is used to forecast the load. Finally, we compared the two methods. Results presented show that the proposed Adding-weighted Largest Lyapunov Exponent Method appears to perform better than the traditional chaotic forecasting arithmetic.