Expanding Existing Solar Irradiance Monitoring Network Using Entropy

Many existing solar irradiance monitoring networks were built particularly for resource assessment purposes; they are often spatially sparse. In order for the networks to handle other increasingly important tasks, such as irradiance forecasting for grid integration, their spatial sparsity must be addressed by adding in new monitoring stations. Optimally expanding these networks using historical information thus becomes an important research topic for engineers. Variability of solar irradiance in space and time can be quantified using statistics such as entropy and covariance. The deployment of the additional monitoring stations should, therefore, utilize these statistics to reduce the variability. More specifically, we aim at maximizing the entropy of the network. A practical difficulty in statistical modeling of solar irradiance is that the data are not ideal. Properties such as stationarity and isotropy are not observed in irradiance random field. We, therefore, focus on hypothesis testing and transformation of the irradiance data, so that the design procedure is statistically justified. We propose the redesign framework in a solar engineering context, using data from 24 irradiance monitoring stations on a tropical island. In the case study, we demonstrate how to find three optimal stations from a pool of 100 potential future monitoring sites.