LAUSR

- A heterogeneous ensemble methods for load forecasting (short-term and mid-term) are proposed here. The proposed approach comprises of a two-level hierarchy of machine learning based methods and classical methods to form the ensemble forecaster, where output of the first-stage forecasters are used as input in the second stage. Artificial Neural Network and Support Vector Regression methods are incorporated in the proposed approach as ML forecasters, whereas Holt’s exponential smoothening and multiple linear regression techniques are included as classical forecasters. The proposed two-level ensemble approach forecasts realistic smart metered data more accurately and efficiently for multiple short-term and mid-term load forecasting scenarios with improved accuracy compared to any individual single stage forecasting methods. The prediction accuracy is shown to improve manifolds for the tested practical system. The proposed model also show improvements compared to existing ensemble based model.