Adaptive software sensor for intelligent control in photovoltaic system optimization

Abstract

• Developed an adaptive observer with a neural and sliding-mode controller. • A sliding mode and neural-based controller achieves 99.9% power efficiency. • Estimation accuracy improved by 4.64% over standard twisting-based observer. • The observer self-adjusts to weather and system parameter changes. • Stability is validated using Lyapunov theory and real-world simulations. This paper addresses the critical challenge of reducing dependence on multiple physical sensors in photovoltaic energy systems, which often leads to increased cost, system complexity, and vulnerability to noise and sensor failures. These issues impact the overall reliability and real-time performance of power extraction and motivate the development of more robust and efficient alternatives. As a solution, an adaptive software sensor is introduced and integrated with a smart maximum power point tracking control strategy for real-time photovoltaic system optimization. The proposed software sensor is designed using an adaptive super-twisting sliding mode observer to estimate internal states, such as inductance current and output voltage. The control strategy is based on artificial neural networks combined with sliding mode control to ensure accurate and stable power tracking under varying environmental conditions. The software sensor’s parameters are automatically adapted in real time to maintain estimation accuracy and robustness. Convergence of the proposed method is analytically verified using Lyapunov theory. Simulation results based on real-world solar irradiance data demonstrate high performance, achieving 99.9% power efficiency and a 4.64% improvement in estimation accuracy compared to the conventional super-twisting observer. These findings confirm the effectiveness of the proposed architecture in enhancing photovoltaic system operation.