Large-scale hydropower projects generate substantial amounts of heterogeneous data dispersed across design, construction, and supervision units. The interoperability among stakeholders is suboptimal due to the heterogeneity of data structures and professional contexts. Consequently, information sharing remains inefficient. Existing studies have typically focused on specific data types or lifecycle stages, lacking a unifying framework to facilitate comprehensive, full-cycle data sharing. To address this issue, this study proposes the development of a data-sharing agent tailored to the needs of hydropower engineering. The proposed agent is designed to accommodate structured, semi-structured, and unstructured data, and it integrates external tools such as time-series databases, knowledge graphs, and text vector databases. This integration enables accurate, on-demand data retrieval. By enhancing the tool-learning capabilities of large language models, the agent bridges data silos, enhances cross-domain collaboration, and lays a solid technical foundation for intelligent construction in complex hydropower projects.

The research commences with a systematic analysis of data-sharing requirements across the full lifecycle of hydropower projects, encompassing time-series monitoring data, technical documentation, and parametric design files. Based on this analysis, a comprehensive agent framework is designed to support multi-modal data interoperability. To ensure its practicality, a supporting tool system is constructed that integrates intelligent modules for database retrieval, knowledge graph querying, and rule-based inference. Furthermore, an action-planning dataset comprising over 4000 samples is developed to train the agent in decision-making and tool invocation. Two versions of the DeepSeek-R1-Distill-Qwen model (1.5B and 7.0B parameters) are fine-tuned using this dataset to enhance structured parameter extraction, multi-step reasoning, and action planning capabilities. To assess performance, a benchmark testing dataset comprising hundreds of real-world business queries derived from hydropower project workflows is established and manually annotated to ensure fairness and reproducibility.

Experimental results demonstrated that the fine-tuned models substantially improved planning and reasoning performance. A comparative analysis revealed that the 1.5B and 7.0B models achieved 270% and 104% improvements in planning accuracy, respectively, compared with their pre-fine-tuning counterparts. On the business query test set, the overall output accuracies were 65.83% and 90.83%, respectively, thereby confirming a significant enhancement in model reliability and practical utility through fine-tuning. Notably, the 7.0B model consistently outperformed the smaller version, highlighting the larger model's capacity to handle complex, multi-step reasoning tasks. A practical deployment of the agent-based data-sharing platform was conducted for a real hydropower project in a representative watershed. Under static and structured data-sharing conditions, the agent maintained an average response time of less than 20s. Conversely, dynamic monitoring scenarios involving high-frequency data streams exhibited average latencies exceeding 30s, with peaks exceeding 60s under intensive analytical loads.

This study proposes a comprehensive framework for constructing a data-sharing agent that effectively addresses critical challenges in current hydropower data-sharing practices, particularly in high-altitude, data-scarce environments. By aligning agent design with engineering-specific requirements and integrating a highly refined large language model with a domain-oriented tool ecosystem, the proposed method significantly enhances the efficiency, intelligence, and semantic interoperability of data sharing. The agent reduces cross-disciplinary access barriers, improves system responsiveness, and supports knowledge-driven decision-making. The results from field applications confirm its considerable potential for practical implementation in intelligent construction platforms. Furthermore, the findings of this study provide a scalable, generalizable technical foundation for the future development of data-driven management and intelligent decision-support systems in complex hydropower projects.

High-altitude hydropower projects present significant challenges owing to harsh environmental conditions, project clustering, limited data availability, and high construction risks. Accurate carbon emission calculations are crucial in such environments to mitigate environmental impacts and promote sustainable development. This study targets the full lifecycle of carbon emissions during intelligent construction in high-altitude hydropower projects.

This study establishes a comprehensive framework for calculating lifecycle carbon emissions tailored to the unique challenges of high-altitude hydropower construction. The methodology covers three primary stages: data collection, model formulation, and real-world implementation. Lifecycle boundaries and emission factors are established for material production, transportation, construction, and operational maintenance. Key emissions are identified based on quality, energy consumption, and cost criteria to build a detailed carbon inventory. To address altitude effects, an adjustment coefficient is derived by correlating field-monitored data with baseline values, accounting for altitude impacts on emission intensities. The carbon emission model incorporates a discrete event simulation (DES) to capture the dynamic characteristics of construction and equipment operations. This model couples static and dynamic elements, applying static calculations to stable phases such as material production and maintenance while using dynamic simulations for variable stages such as transportation and active construction. This DES approach simulates the sequential and interdependent nature of equipment operations, providing an accurate reflection of emission behavior over time. Furthermore, a network of onsite carbon monitoring devices was implemented across different construction sites in a case project, and real-time CO₂ concentration data were collected. These data calibrate and validate emission factors within the model, ensuring accurate altitude-adjusted emission assessments.

The model was applied to the JX hydropower project in a high-altitude region with distinct climatic and geographical challenges. The findings indicated that material production and construction machinery were the largest carbon emitters, accounting for 65.7% and 27.4% of total emissions, respectively. Cement manufacturing was identified as the dominant emission source, emphasizing the need for greener materials and cement production. The DES model revealed that equipment states, such as idling and operation, significantly influence emission intensities, especially under reduced oxygen at high altitudes. By integrating the DES results with real-time monitoring, the model supports precise, responsive emission control strategies. The proposed mitigation measures included adopting cleaner fuels, optimizing equipment idle time, and enhancing operational efficiency through scheduled maintenance. The model reliability was demonstrated by the close alignment of the simulated results with actual onsite measurements.

The developed model offers a structured approach to calculating lifecycle carbon emissions for intelligent hydropower construction in high-altitude regions. By addressing the unique characteristics of such projects, including altitude-induced effects on emission intensities and equipment behavior, the model serves as a reference for emission reduction in future high-altitude hydropower projects. This study advances the understanding and management of emissions in high-altitude construction, underscoring the potential of intelligent construction methods to drive sustainable hydropower development.

Hydropower underground engineering encounters significant safety management challenges owing to overlapping construction activities, diverse process stages, and dynamic resource flows. This involves multidisciplinary safety tasks, such as safety hazard identification and rectification, emergency response, and regulatory compliance checks, which require specialized domain knowledge. In this context, safety management knowledge is intricate, such as expert experience, patterns and characteristics, and management codes, and is dispersed across multimodal data formats, including text, tables, and images. Efficient extraction of these multimodal data sources can significantly enhance data utility and support intelligent safety management. However, owing to the diverse nature of data formats, the complexity of the knowledge system, and the various management scenarios, current research struggles with limited knowledge sources, acquisition difficulties, and poor generalization.

This study proposes a method of constructing a multimodal knowledge graph (KG) for safety management in hydropower underground engineering. (1) A large-scale, high-quality, multisource heterogeneous dataset is built from safety hazard identification and rectification records, regulations, and images. (2) Knowledge modeling employs top-down and bottom-up approaches to define entities, relationships, attributes, and events pertinent to safety management in hydropower underground engineering. (3) The entity and relationship information from text data is obtained using a knowledge extraction method that uses a large language model (LLM) tuned with domain knowledge, enriched by specific examples for each entity type to handle small sample sizes. This approach uses demonstrations to provide the model with prior knowledge. (4) Instance segmentation is used to annotate safety hazard images. The entities identified in the images are then converted into vectors. Image and text data are linked based on semantic similarity. Image data are integrated into the textual KG, enabling the transformation from multimodal data to multimodal knowledge. (5) The multimodal KG is stored in Neo4j, an open-source graph database management system. (6) A scenario-specific knowledge acquisition method addresses the specific needs of safety management scenarios, integrating KG with LLMs to enable retrieval-augmented generation and interpretable knowledge reasoning.

(1) This paper collected more than 120 000 safety hazard records, 30 regulatory documents, and 300 000 images of safety hazards. Leveraging these comprehensive data, this paper constructed a large-scale, high-quality, multisource heterogeneous dataset specifically designed for managing safety in hydropower underground engineering projects. (2) Taking a hydropower underground engineering project as an example, the constructed multimodal KG was applied to intelligent recommendations for safety hazard rectification and compliance checks. (3) The workflow for generating intelligent recommendations for safety hazard rectification measures involved the following steps. After users input safety hazard information, the scene-KG was extracted from the multimodal KG and fed into an LLM to generate appropriate rectification measures. (4) Based on the scene-KG, an inference retrieval method extended neighboring nodes and constructed inference-KG for compliance checks. By integrating inference-KG with an LLM, the system retrieved relevant content from regulatory documents based on user input.

The proposed method effectively extracts and applies domain knowledge from multimodal data in the context of safety management in hydropower underground engineering. It also successfully applies domain knowledge for safety management. The results serve as a reference for transitioning infrastructure construction safety management from a data-driven approach to a knowledge-driven approach.

To address the issue of cracking in the lining structures of deep-buried tunnels, this paper proposes the use of basalt fiber-reinforced concrete (BFRC) to improve the load-bearing capacity of lining structures.

Tension and compression tests were conducted on BFRC specimens with varying volume basalt fiber fractions ranging from 0 to 0.5%. The optimal fiber content was determined, and the concrete damage plasticity (CDP) model parameters for plain and fiber-reinforced concrete with the optimal fiber content were validated. Then, numerical simulations were employed to create an integrated bearing model of surrounding rock-initial support and a secondary lining support model. The use of solid and structural elements and a secondary lining support characteristic curve (SCC) for deep-buried tunnels were obtained, revealing the crack propagation characteristics of the concrete. A quantitative analysis was conducted on the effects of the reinforcement ratio, secondary lining thickness, and fiber-reinforced concrete on the normal and ultimate state load-bearing capacity of the secondary lining.

(1) Compared with those of plain concrete (B0), the maximum increases in the axial tensile strength and splitting tensile strength of BFRC with a fiber volume fraction of 0.2% were 12.81% and 14.79%, respectively. Furthermore, a maximum enhancement of 31.68% in the flexural strength of BFRC was noted when the fiber volume fraction was increased to 0.5%. The optimal fiber content was 0.2%. (2) The stress-strain curve of the BFRC could be fitted using peak compressive strength, peak compressive strain, and compressive shape parameters. The compressive shape parameter values for B0 and B0.2 were 6.50 and 3.00, respectively. The tensile stress-strain curve could be fitted using the peak tensile strength, peak tensile strain, and tensile shape parameter, with tensile shape parameter values for B0 and B0.2 being 3.00 and 1.86, respectively. (3) The CDP parameters for plain concrete and fiber-reinforced concrete accurately simulated the peak tensile and compressive strengths as well as the shapes of the tensile and compressive stress-strain curves. For compressive stress-strain curves, the error between numerical simulation and experimental fitting values at 0.50% compressive strain was 2.48% (2.01%) for B0 (B0.2). For tensile stress-strain curves, the error at 0.04% tensile strain was 4.08% (1.68%) for B0 (B0.2). (4) The SCC curve of the secondary lining exhibited rapid linear growth initially, slow growth in the middle, and a nearly horizontal trend in the later stages with increasing displacement. For class Ⅴ surrounding rock, the secondary lining crack width showed slow linear growth in the initial stage and rapid linear growth after reaching approximately 0.10 mm. Higher reinforcement ratios effectively delayed crack propagation in the early stage, although increasing the reinforcement ratio beyond 0.6% or 0.8% was not economically reasonable. (5) Increases in reinforcement ratio and lining thickness resulted in almost linear increases in the 0.30 mm crack load and ultimate state load-bearing capacity. For every 0.1% increase in the reinforcement ratio, 0.30 mm crack load increased by an average of 5.40%. In addition, for every 0.10 m increase in the secondary lining thickness, 0.30 mm crack load increased by an average of 11.18%. Fiber addition considerably enhanced concrete resistance to crack propagation, especially in the early stages, increasing the initial cracking load by 38.64% and the 0.30 mm crack load by 5.54%.

This study provides theoretical and practical guidance for designing deep-buried tunnel lining structures and serves as a reference for applying fiber-reinforced concrete in secondary lining structures of deep-buried tunnels.

Intelligent ventilation on demand is crucial for ensuring environmental safety in underground caving groups and for the high-quality construction and development of hydropower projects. Ventilation systems for large underground caving groups during construction frequently exhibit complex three-dimensional layouts, different air loads across regions, and dynamic demand under varying regulation conditions.

To achieve spatial node extraction, branch correlation decoupling, and stable joint adjustment of complex flow fields, this paper examines the development characteristics of fluids under construction ventilation in extensive spatial structures. It demonstrates the necessity of constructing a graph structure based on the ventilation flow characteristics for analyzing and adjusting ventilation system parameters. The regional modeling theory is discussed, detailing the principles and methods of node extraction for one-dimensional tube bundle fluids (network) and three-dimensional spatial flow field elements (field). Among these, the area where fluid parameter information changes along the main airflow direction employs network node extraction, while the regions with multi-directional complex flow paths utilize the three-dimensional field node extraction method. Virtual branches address the network-field coupling problem, utilizing the nodal pressure approach. This method treats the nodal pressure as the unknown variable and airflow deviation as the assessment criterion. Nodes with known pressure values serve as reference nodes for solving the pressure at all network nodes, and are further assigned to field simulation boundaries. By numerically simulating the three-dimensional spatial flow field, the virtual branch air flow rates are iteratively fed back into the air network calculation for a coupled solution. This paper also introduces the node-property-edge triplet, which effectively reflects the structure, performance, and behavioral characteristics of nodes. Furthermore, to optimize the ventilation coordination efficiency, a hypergraph structure for joint adjustment, with edges as the analysis object, displays the coupling interactions between the ventilation branches and loops. Considering the joint adjustment sensitivity, an optimal resistance control method is proposed, which involves constructing target and response node sets, setting response efficiency constraints, and optimizing to form a ventilation adjustment plan. An intelligent ventilation coordination platform integrates the resistance control model of coupling interactions, including modules for network design, ventilation design, field-network integration, loop generation, and optimization analysis. Within this framework, the network design module is dedicated to reconstructing the physical model of the ventilation system, while the ventilation design and field-network integration modules are used to assign basic fluid characteristic parameters of ventilation to the established model. The loop generation and optimization analysis modules are employed for solving the overall wind network parameters, including air volume, air pressure, and wind resistance.

The field-network coupling method using nodal pressure eliminated the need for loop identification and effectively addressed the interdependent coupling between network nodes and flow field boundaries. The intelligent ventilation coordination platform was integrated with online environmental monitoring devices to automatically gather critical ventilation environment parameters, thereby enabling real-time calculations of the ventilation system based on environmental monitoring data and providing 3D visualization and early warning capabilities.

The ventilation design parameters of an engineering project are used to implement targeted air volume control deployment. The integrated control system exhibits high responsiveness. On the premise that the air volume of each unit meets the threshold requirements, the air volume adjustment efficiency of the target unit and the overall stability of the air distribution network can always fulfill the specified requirements. The results indicate a timely and stable system response and can provide a reference for similar projects.

Anchors are typical mass concrete structures found in large bridges, characterized by large structural sizes, complex boundary conditions with irregular shapes, low reinforcement ratios, high crack resistance requirements, and challenges in temperature control and crack prevention. The development of an adaptive, intelligent cooling control method and system is crucial for crack prevention and improving concrete pouring quality.

This paper proposes an intelligent cooling control method for bridge anchorage, including: (1) The basic control principles of heat balance of supply and use, accurate control, and online warning. (2) A fundamental intelligent control strategy involving real thermal field simulation and a temperature-flow coupling control algorithm. The combined influence of temperature and flow is considered when predicting the cooling system parameters. This study uses a hybrid approach involving a long short-term memory neural network (LSTM) and proportional integral derivative (PID) control algorithms to predict the future water flow rate based on the current concrete and cooling system state parameters, facilitating the temperature-to-flow mapping. (3) A "multiple terminal-edge computing-cloud storage" control model is implemented, which incorporates edge computing within the control cabinet, providing localized endpoint services to improve data transmission performance, ensure real-time processing, and reduce latency. Cloud computing uses machine learning to provide instructions for adjusting temperature and flow rates based on the deviations between the actual and target temperature control curves. Furthermore, fault recognition and rapid diagnosis functions are also implemented. Intelligent cooling control equipments and code platforms are developed for realizing online perception, real analysis, feedback control, remote diagnostics, and early warning systems for the cooling process. The system comprises water supply, reversing, control and heat exchange subsystems, and a multiterminal software platform based on WeChat and the web.

This paper adopted simulation, equipment development, and field application methods based on the Longmen Bridge project. Real temperature field simulation calculations were conducted, the temperature distribution during the cooling process was analyzed, and the impact of heat transfer from the upper layer of concrete, as well as the design of cooling pipes, was optimized. Parameters such as water temperature, water flow, concrete temperature, and temperature gradient were analyzed. Furthermore, as part of a long-term temperature monitoring process, the impact of heat transfer from the upper layer of concrete was assessed to reduce the temperature difference between layers. A personalized water-cooling strategy was proposed, and the timing of the water supply was adjusted.

The established temperature-flow coupling control algorithm, model, equipment, and platform achieve real-time monitoring, analysis, control, continuous optimization, and early warning of water-cooling information online and remotely. The study results are successfully applied to the west anchorage of the Longmen Bridge. No temperature cracks are observed on the bridge site, which reduce manpower and water consumption. The results can be used as a design and construction reference for thermal cracking control in similar projects.