A Hybrid CNN-LSTM Framework and Infrared Image Processing for Solar Irradiance Forecasting

Accurate solar irradiance forecasting is crucial for optimizing photovoltaic power generation, especially given the intermittent nature of solar energy. This paper proposes a novel framework for short-term solar irradiance forecasting using low-resolution infrared (IR) sky images. By leveraging bicubic interpolation and OpenCV colormaps, the framework enhances the quality of IR images, providing better inputs to a Convolutional Neural Network (CNN)-based nowcasting model, which is followed by a Long Short-Term Memory (LSTM) model for univariate solar forecasting. Key results indicate that the proposed method reduces the Root Mean Squared Error (RMSE) from 47.24 W/m<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> to 33.51 W/m<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>. Extensive evaluations demonstrate that image processing techniques significantly improve forecasting accuracy compared to baseline models using raw IR images.