A Practical Real-Time Observer-Based Radiation Prediction Algorithm for Solar Plants

Abstract

The global transition toward clean energy has intensified interest in solar power, especially in regions with favorable geographical conditions. Despite the rapid development and deployment of solar plants, operational challenges remain, particularly in optimizing energy conversion in real time. This paper proposes a practical real-time solar radiation prediction model designed to enhance the performance of solar plants by forecasting available energy, thereby improving control during the energy conversion process. To this aim, an autonomous nonlinear dynamical model with an unknown drift function is considered. A Group Method of Data Handling (GMDH)-based identification approach, supported by a comprehensive experimental dataset, is employed to estimate the drift function and confirm the feasibility of the model. Once the nonlinear model is validated, a theoretical framework is developed to enable adaptive estimation of the model's states and parameters, eliminating the need for offline identification. Experimental results across multiple scenarios demonstrate the model's effectiveness in accurately identifying unknown parameters and state variables under different environmental conditions, geographic locations, and challenging cases such as partial shading. These results highlight the practical potential of the proposed method for improving real-time control and energy efficiency in solar plant operations. © 2025 The Author(s). Energy Science & Engineering published by Society of Chemical Industry and John Wiley & Sons Ltd.