Local Partial Least Squares Multi-Step Model for Short-Term Load Forecasting(Modelling, Systems and Simulation,<Special Section>Nonlinear Theory and its Applications)

概要

論文の詳細を見る
The multi-step prediction model based on partial least squares (PLS) is established to predict short-term load series with high embedding dimension in this paper, which refrains from cumulative error with local single-step linear model, and can cope with the multi-collinearity in the reconstructed phase space. In the model, PLS is used to model the dynamic evolution between the phase points and the corresponding future points. With research on the PLS theory, the model algorithm is put forward. Finally, the actual load series are used to test this model, and the results show that the model plays well in chaotic time series prediction, even if the embedding dimension is selected a big value.
社団法人電子情報通信学会の論文
2006-10-01