The rapid growth of rooftop solar power (RSP) systems in Vietnam and many other countries has created an urgent demand for reliable real-time monitoring and power estimation solutions, particularly for small-and medium-scale installations. However, existing commercial systems often involve high deployment costs, strong dependence on network infrastructure, and limited flexibility in remote areas. This paper proposes a low-cost AIoT-based monitoring and power estimation system that integrates Internet of Things (IoT) technologies with machine learning models to improve scalability, resilience, and adaptability under real-world conditions. The system is implemented on low-cost hardware using a Raspberry Pi as an edge computing node, enabling synchronized acquisition of electrical data from inverters and environmental data from sensors via industrial communication protocols. To estimate the DC power output, two machine learning models including eXtreme Gradient Boosting (XGBoost) and Support Vector Regression (SVR) are trained using real measurement data collected from a 1.2 MW rooftop PV system in Da Nang, Vietnam. The estimated results are compared with actual measurements to evaluate model performance under varying weather conditions. Experimental results show that both models achieve high accuracy, while XGBoost demonstrates superior performance under rapidly fluctuating conditions. In addition, the system integrates a real-time alert mechanism to detect abnormal deviations between estimated and measured power, enabling early fault identification. By combining edge computing and cloud-based services through the open-source MING stack, the proposed solution ensures low latency, stable operation during network interruptions, and efficient remote monitoring. Overall, the proposed AIoT framework provides a practical, cost-effective, and scalable solution for monitoring and optimizing rooftop PV systems.
Nguyen et al. (Wed,) studied this question.