Design and Evaluation of an Affordable Open-Source All-Sky Camera for Measuring Diffuse Sky Radiance

The growing reliance on photovoltaic energy requires detailed solar irradiance measurements to ensure grid stability. However, current methods are limited by the established model of solar irradiance, which reduces the complex, non-uniform solar radiation down-welling from the entire sky dome to a direct irradiance component and a single value for diffuse irradiance, leading to systematic errors. To improve accuracy, we must move beyond single-point measurements of diffuse irradiance and map the entire sky radiant distribution instead. All-Sky Imagers have been shown as a viable technology to capture angularly and spectrally resolved diffuse sky radiance. Currently, the lack of standardized, low-cost, and open-source hardware and software for all-sky imaging hampers research and reproducibility. This work addresses these issues by developing and validating a complete system for measuring pixel-wise diffuse sky radiance in three spectral bands in the visible spectrum. A low-cost all-sky imager prototype was designed and built for approximately 304€ using a Raspberry Pi 4, a SONY IMX708 image sensor, and a fisheye lens. A corresponding open-source software pipeline was developed entirely in Python. The methodology consists of a two-step geometric calibration to map image pixels to unique directions in the sky and a radiometric calibration that relates the sensor’s signal to physical radiance values(W m−2 sr−1) using the libRadtran radiative transfer model as a reference. The system successfully produced high-resolution, pixel-wise diffuse sky radiance maps. The geometric calibration was validated, and the radiometric analysis established a strong, monotonic relationship (Spearman correlation ≈ 0.95) between the sensor’s signal and simulated radiance. However, the prototype’s physical implementation failed partially: the custom-built enclosure proved mechanically unstable due to thermal expansion, causing a daily shift in the field of view that invalidates the geometric calibration over time. In conclusion, this work delivers a validated blueprint for a low-cost all-sky imager, confirming the feasibility of the approach. While the primary contribution is the end-to-end methodology for radiance map generation, the most significant shortcoming is the physical prototype’s failure due to insufficient stability and reproducibility. A complete mechanical redesign of the enclosure is, therefore, the most critical area for future work.