Novel Semi-Automatic Method to Optimize Multi-Lamp High Flux Solar Simulators

Abstract

For multi-lamp high flux solar simulators (HFSS), it is often difficult to obtain a required flux distribution by manipulating the lamp position of multiple lamps at once. Each lamp has three degree of freedom. Thus manual optimization can be tedious for human operators. Thus, this project aims to create a semi-automatic method to determine the optimal location of the lamps to give the required flux distribution. A convolutional neural network is used to develop a mathematical model that performs the above function. At the same time, an automated method to collect data from the HFSS was devised. Furthermore, an in-house algorithm to characterize the irradiance was developed. Since large amount of data was required, an optical simulator called TracePro was used to generate the data for training as well as validation. This project serves as proof of concept of using machine learning to optimize HFSS. In the long term, the proposed methodology is expected to facilitate initial deployment of the HFSS. It will also assist on the dynamic control of reactor conditions i.e. emulating variable overcast or daily sunlight variability.