Exerting huge economic and social benefits, the construction and operation of water conservancy projects have an impact on the ecological environment of upstream and downstream reaches. However, it is also an effective way to regulate and improve the ecological environment by rationally optimizing the water storage and release process of the reservoir and carrying out targeted ecological management of the reservoir. From the perspective of ecological flow constraints, through data investigation and analysis, the appropriate flow ranges of the Yangtze River main stream conducive to the natural reproduction of the four major Chinese carps, Chinese sturgeon and the control of estuary salt tide intrusion were determined. Then, the optimal scheduling model of the Three Gorges Reservoir was established considering the downstream ecological flow constraints. Multi-objective reservoir ecological dispatching model is emphasized. The results show that the ecological flow of the lower reaches of the Three Gorges Reservoir was basically guaranteed during the wet season from May to September, but the ecological water shortage was prone to occur in different degrees in other seasons, especially in the dry years. For the optimal scheduling, the power generation and the monthly maximum ecological water shortage rate showed a positive relationship, indicating that the two scheduling objectives could not reach the optimum simultaneously. The power generation of the optimal scheduling in a typical dry year was between 80.2 billion and 83.5 billion kW·h, and the monthly maximum ecological water shortage rate was between 24% and 48%. The optimal scheme obtained by multi-objective dispatching can basically coordinate the ecological and power generation needs. On the basis of the downstream ecological flow needs, the power generation benefits increases by 4.07% compared to the design scheme.

In view of the fact that the reservoirs of medium and small hydropower stations usually operate at a lower level during the flood season because of the smaller regulating reservoir capacity, short decision time for flood regulation, maximum head of the unit and inundation limitation of the upstream reservoir area, etc., the flood resource utilization benefit is not fully utilized during the flood season regardless of the magnitude of the flood. The multi-objective risk analysis model for flood operation and scheduling of medium and small hydropower reservoirs is established with the objective of maximizing the power generation benefit and minimizing the flood risk, and the solution method for the optimal flood level is given when the forecasted incoming flood flow is in different ranges. The results show that when the reservoir faces different levels of floods, the optimal operation level can be obtained by coordinating the benefits and risks to increase the power generation benefits during small floods and reduce the flood losses during large floods, which provides a reference for the flood operation and scheduling of medium and small hydropower reservoirs under changing environment.

Limited by hydrometeorological data, flood forecasting in ungauged basins still faces challenges. Parameter regionalization is a common method to solve this problem. The machine learning model has the characteristics of simple modeling and convenient use compared with the traditional flood forecasting model. Taking the West Plain of Nansihu Lake in Shandong Province as the research area, referencing the idea of hydrological regional synthesis, this paper synthesizes the data of 40 floods in 8 watersheds from 2010 to 2021, and builds a regionalized flood forecasting model based on Long Short-Term Memory (LSTM). The results show that the regionalized flood forecasting model can simulate the actual flood process well, the relative error of flood peak in both the training set and the testing set are less than 10%, and the Nash-Sutcliffe efficiency coefficients are all greater than 0.9; In the 15 h forecast period, the regionalized flood forecasting model has higher forecasting accuracy, and when the forecast period is more than 15 h, the forecast accuracy of the model decreases.

An accurate hydrological model with strong adaptability is crucial for predicting floods and preventing disasters. The antecedent soil moisture and rain intensity factors are considered to improve the SCS-CN runoff generation model. Then the new SCS-CN hydrological model is developed by adding the watershed concentration module. From the perspectives of different climate zones and magnitude floods, the new model is applied in four basins to flow simulation, which include semiarid basins of Hancun and Macun, and humid basins of Shenglihe and Gongcheng. The results demonstrate that new SCS-CN hydrological model has good applicability in both semi-arid and humid basins, with better performance in semi-arid basins. Generally, the model shows better accuracy in simulating floods of different magnitudes, especially for small and medium floods.

From the objectives, variables and relations of the time-varying linear confluence model, it can be seen that the model has some constraints such as static parameters and the inability to consider the influence of interval runoff, and there are significant defects in the application of flood prediction. Therefore, the time-varying linear confluence model is improved by combining the chaotic mapping with the rich model to solve the diversity and adding the influence operator to replace the interval runoff. Using the measured runoff data of Maoergai Hydropower Station in the Heishui River basin for many years, and taking the forecast process, flood volume, flood peak and peak time as the evaluation index, the application analysis of the improved time-varying linear confluence model is carried out. The results show that the overall prediction qualification rate of the improved model is increased by 9.13%, and the certainty coefficient is increased by 0.25, which expands the reliability and practicability of the application of the time-varying linear confluence model.

In order to enhance the real-time flood forecasting accuracy in the Linyi River Basin, a TOPKAPI grid model was developed based on the underlying surface characteristics of the Linyi River Basin. The TOPKAPI model simulation results were corrected at different lead times using BP neural networks and LSTM models. Furthermore, a stacking approach was applied, employing the Transformer model as a secondary learning tool to refine the corrections made by BP and LSTM. The results indicate that after real-time correction with the BP and LSTM models, the improvement of the simulation accuracy of the TOPKAPI model is obvious, with better correction results for shorter lead times. Following the stacking method for secondary learning, the correction results is the best, effectively enhancing the flood forecasting accuracy in the Linyi River Basin.

The regulation and storage capacity of a river channel plays an important role in the safety of drainage in polder areas. In order to study the factors related to the actual effective regulation and storage capacity of a river channel in polder areas, this paper adopts the unsteady flow model and the water balance method to calculate the regulation and storage capacity of a single river channel. The calculation results of the two methods show that the main influencing factors of the regulation and storage capacity of a river channel are the water depth corresponding to the starting regulation level, the roughness of the river channel, the length of the river channel and the cross-section morphology, etc. For the backbone drainage channel with large discharge, the actual effective regulation and storage capacity is the storage capacity from above the water surface line corresponding to the designed discharge to the maximum water level. The research results have certain reference value for the planning and design of drainage in polder areas.

In order to explore the causes of urban waterlogging, taking Guangzhou as an example, a drainage model was established based on SWMM. A waterlogging cause analysis method based on real-time monitoring data was proposed. Combining with the real-time monitoring data, the water-collecting capacity of water-logging points, the flow capacity of pipeline and the runoff generated by rainstorm were calculated. At the same time, the corresponding waterlogging measures were given according to different waterlogging causes. The results show that 75% of waterlogging cases involved insufficient surface drainage capacity, 57.7% of waterlogging cases involved watercourse top support, and 44.2% of waterlogging cases involved excessive rainfall intensity and insufficient pipe drainage capacity.

Aiming at the shortcomings of slow convergence and easily falling into the local minimum in the learning process for the traditional BP neural network, particle swarm algorithm with fast convergence speed and strong global optimization ability was introduced so as to establish the PSO-BP model. Taking the seepage monitoring data of an earth and rockfill dam as an example, the seepage was predicted. Compared with the prediction model, the BP model and the traditional statistical regression model, the results show that the PSO-BP model has a higher goodness-of-fit and convergence.

Conventional dam displacement monitoring methods are often associated with large errors and low efficiency. Manual monitoring methods cannot provide continuous real-time monitoring, while automated monitoring methods, such as total station robot and GNSS, are affected by weather and have limited accuracy for vertical displacement. To overcome these shortcomings, this study proposes a new intelligent monitoring method for dam displacement based on machine vision. The method utilizes the internet of things and intelligent disaster recognition algorithm to convert picture data into deformation data, enabling ultra-high precision non-contact real-time measurement of the dam. The monitoring system was tested at Lianghui Reservoir, the results demonstrate that the operation of monitoring system is stable, and the horizontal and vertical monitoring accuracy are both 1.5 mm. The proposed method has the potential to be widely applied in other water conservancy projects for surface displacement monitoring.