Short-Term Load Forecasting Based on Modified Locally Linear Embedding Using Multiple Weights

This paper proposes a novel short-term load forecasting method by using a manifold learning algorithm—modified local linear embedding (MLLE). By taking the 24h load data as a 24-dimensional data set, the intrinsic properties and overall geometric laws of high-dimensional data were extracted by manifold learning method MLLE to reduce dimension, and the load prediction was carried out in low-dimensional space. Finally, the 24-hour load prediction was obtained by using MLLE reconstruction. Simulation results show that that the method proposed in this paper has high prediction accuracy and stable prediction results.