PV Power Prediction via Residual Correction and Hierarchical Multiple Buildings Energy Management

Abstract

This paper deals with photovoltaic (PV) power prediction and building energy management system (BEMS) in multiple buildings. First, we consider PV power prediction. We predict PV power by using k-means method and SVR. However, there is an error between prediction and actual generation. In order to solve this issue, we estimate prediction error in advance by markov process and revise the prediction. And, by applying copula to calculated point prediction, calculate interval prediction. Second, we consider 2 level BEMS. At first level, in each building, power consumption is determined considering comfort. At second level, in order to reduce total cost in all buildings, battery scheduling is calculated with robustness by applying scenario robust theory. And, at the end of each section, simulation results show the effectiveness of proposed prediction method and energy management method.