SVR based PV models for MPC based energy flow management

No Thumbnail Available
Boegli, M.
Stauffer, Y.
The ever increasing penetration of renewable production can be used to lower the operational costs of households and districts. However, given its intermittent nature predicting their production in order to use it efficiently is mandatory. This article will discuss how photovoltaic production prediction can be addressed. Prediction algorithms based on support vector regression (SVR) technique are developed and compared to a classical persistence model benchmark. Prediction errors between 7 to 12% for SVR algorithms with different forecast solar irradiance accuracies are obtained, which is twice better compared to a classical persistence model (PM) benchmark. These models were integrated in a model predictive based battery management system in the scope of the European project AMBASSADOR. (C) 2017 The Authors. Published by Elsevier Ltd.
Publication Reference
in Cisbat 2017 International Conference Future Buildings and Districts - Energy Efficiency from Nano to Urban Scale. vol. 122 (Issue), J. L. Scartezzini, Ed., ed Amsterdam: Elsevier Science Bv, 2017, pp. 133-138.