In-Situ UNet-Based Heliostat Beam Characterization Method for Precise Flux Calculation Using the Camera-Target Method

Abstract

Concentrated Solar Power (CSP) utilizing solar tower power plants exhibits significant potential for dispatchable sustainable power generation and fuel production, by concentrating up to hundred thousand focal spots of individual mirrors onto one surface. The irradiance on the receiver is a critical parameter for power plant operators, not only to prevent component-damaging temperature peaks and gradients but also to optimize the distribution of heliostat’s focal spots. Therefore, recent approaches aim to exploit images of isolated focal spots on a Lambertian target, taken on a day-to-day basis in most commercial power plants, to achieve more accurate flux predictions. State-of-the-art methods for flux measurements require additional equipment and are mostly time-intensive. This study introduces a novel methodology to derive single heliostat irradiance profiles using simple digital camera images, employing an image-to-image neural network. The model we propose is capable of generating accurate, high-resolution flux distributions from single images. It uniquely accounts for the target surface’s irregularities and reflectivity as well as background irradiation, factors critical for precise measurements of individual focal spots. By incorporating these factors, our methodology offers a new scaling approach for the measured flux density based on a calculated total flux, obviating the need for additional equipment such as flux gauges or moving bars. This enables to effortlessly be integrated into the daily calibration routine in-situ using the cameratarget method. This seamless integration not only supports existing flux prediction methodologies but also paves the way for future enhancements by potentially improving the accuracy of overall flux prediction.