The safety hazards associated with hydropower construction are diverse and often occur in complex and variable spatial contexts. Human visual assessments based on experience are prone to cognitive and psychological biases. Existing studies face key limitations, including unimodal models failing to capture cross-modal hazard features, the limited generalization ability of small models, and the poor domain adaptability of general-purpose pre-trained models. To address these challenges, this study establishes the first multimodal image-text dataset for safety hazards in hydropower engineering. By leveraging the Qwen2.5-VL model, we implement efficient domain adaptation through LoRA and instruction tuning. A multimodal large model fine-tuned for the intelligent recognition of safety hazards in hydropower construction was proposed. Comparative experiments across hazard types, modalities, and model architectures reveal that: (1) data imbalance has a limited impact on performance differences across hazard types, (2) textual descriptions generally convey more critical hazard-related information than visual features, and (3) domain-specific fine-tuning and effective multimodal fusion are identified as key factors in enhancing hazard recognition performance.

This paper collected relevant research literatures on the smart water conservancy in the databases of China National Knowledge Infrastructure (CNKI) and Web of Science (WOS) from 2000 to 2021. By using the VOSviewer, CiteSpace and other software, this study built various knowledge maps for the time series distribution of literatures in the field of smart water conservancy, publishing institutions, and evolution of research hotspots, to analyze the current progress of smart water conservancy research. The results show that the literature amount of smart water conservancy is increasing year by year, but there is a significant gap between the CNKI database and the WOS database, and core research institutions have been formed in the field of smart water conservancy making important contributions to the frontier development. The CNKI database focuses on the construction of digital watershed and smart water conservancy framework by basin as a unit, while the WOS database focuses on researches from the perspective of geography and earth. Both of them build the platforms for smart water conservancy based on the Internet of Things and deep learning.

To deeply and comprehensively explore the mechanism of the risk caused by dam break accident, this study proposes a calculation method of the risk caused by dam break accident through combining the mining of historical dam break accident cases and Bayesian network. Based on a large number of historical cases of dam break accidents at home and abroad, 24Model is used to identify and extract the causes and chain of dam break accidents. The topology structure caused by dam break accident is constructed, and the probability of dam break is calculated by Bayesian forward causal reasoning, and the mechanism of dam break is analyzed by reverse diagnostic reasoning. Based on the Bayesian sensitivity analysis, the key risk factors affecting dam failure are explored. The results show that in terms of human factors, the proportion of gate control problems is high, while in terms of management factors, construction problems, operation and maintenance management defects, and design problems are important indirect causes of dam break. Flood overtopping and seepage erosion/piping are the main risk factors leading to dam failure.