On improving solar forecasts with enhanced IR imaging and CNNs

Renewable energy such as solar power has been widely adopted lately due to its sustainability and cost- effectiveness. However, due to the intermittent nature of photovoltaic systems, accurate solar forecasting is vital for efficient integration into the grid. In this paper, we propose a novel framework for effective solar irradiance estimation from infrared sky images. The proposed method utilizes contrast-limited adaptive Histogram Equalization (CLAHE) and OpenCV colormaps for pre-processing low-contrast and low-resolution infrared sky images. Additionally, it introduces an Inception module-based CNN-Regression model to get rich feature information and avoid feature loss. We perform a detailed comprehensive study to evaluate the effectiveness of the proposed method. While the baseline CNN regression model (without enhancement and Inception module) achieves an RMSE of 81.30W/m 2, employing the proposed technique reduces the loss to an impressive 47.85W/m 2. With a novel approach to address the existing issues of infrared sky images, the research advances the field of solar power estimation and forecasting for seamless integration into grids and residential settings.