| The precise prediction of power generation and electricity load is critical to the efficient operation of state of the art energy management schemes, especially under heterogeneous, dynamic, and changing environments. The rise in the usage of renewable power generation, coupled with its high dependency on weather related factors, is a major source of uncertainty, limiting the applicability of traditional prediction schemes. In this paper, a unified prediction model for photovoltaic power generation and short term electricity load demands is proposed, based on data fusion techniques. The framework comprises a hybrid model for power generation, based on a combination of physics based models for photovoltaic power generation, coupled with machine learning based residual correction models. The other is an ensemble load forecasting method employing meta model fusion. The proposed models are experimentally validated on real power generation data, based on a solar park in the Thessalia region, Greece, and the results prove the superiority of the proposed method in dynamic and extreme situations. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.