Optimal Irradiance Sensor Placement for Photovoltaic Systems Using Mutual Information Based Greedy Algorithm in Gaussian Process

Abstract

This paper proposes the application of the mutual information criteria based greedy algorithm to place the irradiance sensor for photovoltaic (PV) systems. There is little information about irradiance sensor placement for PV systems in literature because of the complexity caused by the variability of the irradiance distribution. Existing methods in the literature are not able to provide good accuracy due to either their experience based characteristic or the low resolution in the satellite data used to determine the irradiance distribution. In this work, to get the near optimal sensor placement, that is, the best $b$ locations out of $n$ possible sensor locations, a mutual information based greedy algorithm is used to maximize the mutual information increase (MII) in the unsensed locations. The kriging interpolation technique is then used to predict the irradiance values at these unsensed locations. The effectiveness of the greedy algorithm based on the mutual information criteria is verified using experimental datasets measured at two different locations, namely at Nanyang Technological University, Singapore and NREL Oahu, Hawaii. The results show that the proposed method can provide a near optimal sensor placement which provides close performance to the minimum average root mean square error (RMSE) of the irradiance values at the unsensed locations achieved using an exhaustive search.