Light energy concentrating systems such as High Flux Solar Simulators (HFSS) offer notable advantages in renewable energy research. Despite their expanding usage, they face some operational challenges that provide opportunities for new designs and further exploration. Accordingly, this study investigated the development and performance of a machine learning model based on convolutional neural networks (CNNs) for HFSS operation and control. Specifically, the hypothesis proposed is that a CNN can predict three HFSS parameters (the relative x, y, z position of the light source) using imaging and computer vision techniques with an accuracy equal to or better than the operator. First, the HFSS output was characterized in detail to set the overall study expectations and serve as the baseline metric. Then, the optical modelling of the HFSS using Monte Carlo Ray Tracing was employed to generate more than 2,500 images for the training and validation of the CNN. Consequently, the hypothesis was validated as the CNN accurately predicted the source position within 0.07, 0.11, and 0.07 mm in the x-, y-, and z-directions, respectively (a Euclidean distance of similar to 0.249 mm). This is comparable to the mechanical system's accuracy, which allows the positioning of the light source with a Euclidian distance of approximately 0.24 mm. Remarkably, this achievement allows for full automation and better control of light concentrating facilities and the development of more innovative energy harvesting systems.