Large-scale photovoltaic power plants located in highland mountainous areas are vulnerable to landslides due to extreme rainfall, posing a significant threat to the normal operation of photovoltaic power plants. However, limited research has been conducted on landslide risk assessment specifically tailored to large photovoltaic power plants, with most studies focusing on static assessments that lack long-term sustainability in risk assessment and prediction. In this paper, a dynamic study on landslide risk at a large photovoltaic power plant project under extreme rainfall conditions is conducted. Firstly, the factors in landslide susceptibility assessment based on typical landslide characteristics in the study area are selected and an assessment index database using mapping units to extract the relevant factors is established. Subsequently, the ANP-FBN model is employed to evaluate the landslide susceptibility of large photovoltaic power plant sites. Furthermore, an assessment index system for landslide hazard vulnerability is developed by considering population, economic, and material vulnerabilities, and the AHP method is adapted to assess landslides vulnerability in the study area. Finally, the landslide rainfall threshold with the susceptibility and vulnerability assessment results are coupled to achieve a dynamic assessment of landslide hazard risk at large photovoltaic power plant sites under extreme rainfall conditions. The findings highlight that the central valley and the eastern steep slope of the study area are the primary “high” and “very high” risk areas. Moreover, with the increase in rainfall duration, the risk level of landslide hazards in the study area also rises.

Cemented soil offers advantages over conventional concrete, including reduced carbon emissions, cost-effectiveness, and recyclability, making it suitable for foundation reinforcement and excavation support. However, reliable health monitoring of cemented soil structures remains challenging. This study introduces multi-walled carbon nanotubes (MWCNTs) and nano-magnetite (NM) as conductive fillers to develop self-sensing cemented soil composites for real-time monitoring of pile foundations. The electrical conductivity, piezoresistive response, and dynamic sensing behavior of composites prepared with different doping strategies were systematically evaluated through model pile loading tests and microstructural analyses. Results indicate that MWCNTs enhance electro-mechanical coupling by forming conductive networks, while NM primarily improves early mechanical properties and densifies the matrix. Combined doping mitigates nanotube agglomeration, optimizes conductive pathways, and enhances microstructural integrity. Under cyclic compression, composites achieved a maximum fractional change in resistivity (–35%) with strong linear correlation to strain ( R ² > 0.90), maintaining stable responses across repeated loading cycles. Single-pile horizontal loading tests further demonstrated effective signal output across monitoring depths, synchronized with strain fluctuations, and preserved even in deeply buried regions, indicating robust signal penetration and multi-layer sensing capacity. Moreover, the conductive network remained stable under large deformations and micro-crack evolution, confirming high adaptability and fault tolerance. These findings provide strong evidence that co-doped MWCNTs and NM cemented soil composites are reliable and scalable materials for real-time health monitoring of pile structures under practical engineering conditions.