Estimation of building energy consumption using weather information derived from photovoltaic power plants

Abstract

Abstract Photovoltaic power must be measured for billing purposes and to provide power injection information. To emphasise the importance of weather information derived from photovoltaic power data, we consider how building energy consumption is estimated. Photovoltaic power can be treated as an input to an energy consumption model rather than weather information (solar insolation, temperature, and/or relative humidity). We use a partial, mutual information algorithm for selection of the input variables required by a building consumption model; the data are derived from adjacent photovoltaic power stations. When weather information imparted by photovoltaic power is inadequate, the accuracy of energy consumption estimations can be improved by combining an empirical mode decomposition algorithm and an extreme-learning machine algorithm. Our energy consumption estimations, based on partial mutual information, empirical mode decomposition, and use of an extreme-learning machine, were verified using real data from Beijing and Guangzhou, China. The simulations show that the precision of estimation can be increased by fully exploiting the interdependence of photovoltaic power and building energy consumption.