When the general segmentation model is applied to the apparent cracks in the dam face concrete, the network suffers the problem of depth increasing that leads to excessive model parameters and certain loss of effective crack features. To reduce network memory occupation and feature loss, this paper develops a lightweight crack segmentation method based on a convolutional neural network. The network adopts an encoding-decoding structure, and uses a depth-separable convolution module and a lightweight feature extraction module to construct a cascade encoder; it is equipped with a decoder to fuse cross-scale information in the second stage of the encoder and to reconstruct the pixel-level geometric information lost in feature extraction to improve the accuracy of network segmentation. The experimental results show the model size of the network trained on the crack dataset of dam face concrete is 10.8 MB or a size reduction of 90.8% from U-Net, with its PA of 73.3% and IoU of 85.4%. The results verify the network is feasible in dam face crack segmentation and useful for improving the efficiency of dam face detection and maintenance.

Based on the fifth generation atmospheric reanalysis dataset ERA5 of the European Centre for Medium Range Weather Forecasts (ECMWF), we define two types of hot droughts or compound drought-hot events－simultaneous occurrences of meteorological drought/agricultural drought and high temperature. We examine the evolution of such an event occurred in the Yangtze River basin in the summer of 2022, and evaluate the variations in its several characteristics such as duration and spatial coverage. The results show that this hot drought began in June, became most severe in August, and weakened in September; its spatial scale varied significantly, starting from the middle and lower reaches, gradually expanding to the whole basin, and reducing to the middle and lower reaches by September. And compared with typical events in historical periods, its characteristic values were the largest. We find a significant increase in the characteristic values of the two types of compound drought-hot events in July and August from 1979 to 2022. The results deepen our understanding of the hot droughts and extreme events in a river basin and can be useful for coping with extremes under global warming.

To explore the effect of deep learning algorithms combined with the multi-source data fusion method in watershed runoff prediction, a bidirectional Long Short-Term Memory (LSTM) neural network model and a data fusion algorithm of the ensemble Kalman filter are combined to construct runoff prediction models for five watersheds in the upper Hanjiang River. These models are verified using long-series hydrometeorological datasets from the study area and atmospheric circulation factor datasets. The results show that in the same prediction period, the models improve the prediction indexes and better capture the extreme values of runoff series in comparison with the traditional LSTM model. After the data fusion algorithm is used to join the atmospheric circulation factor datasets, the evaluation indexes of different watersheds can be further improved, and their time variations are more stable with a longer forecasting period. These prediction models are effective in improving deep learning-based runoff predictions.

Climate change is closely related to energy-power supply and demand. In the summer of 2022, the upper reaches of the Yangtze River experienced three worst cases: the highest temperature, longest continuous hot days, and lowest rainfall in the same period in history. This resulted in a daily power shortage of 17 million kW and 370 million kW?h in Sichuan, a major hydropower province, and imposed a significant impact on its social and economic development and people’s livelihood. To guarantee energy and power safety, it is of great significance to establish power planning mechanism and some countermeasures for power supply guarantee in extreme weather. This paper presents an analysis on the impacts of extreme weather in the 2022 summer on the power supply guarantee in Sichuan, and examines the shortcomings of previous electric power development. We suggest certain countermeasures for the period of power transformation－such as water-wind-solar-thermal energy complementarity, and a coordinated development of power supply and power grid.

Offshore Floating Photo Voltaic (FPV) is an effective way to deal with the contradiction between photovoltaic development and land resources. However, under the action of sea waves, the photovoltaic panels on a FPV structure always oscillate with wave motion, which makes their angles to the sun change constantly and imposes a great impact on power generation. In this paper, a formula is derived for calculating the radiation energy of the panels under regular sea waves, and the concept of time-equal-dip angle is summarized and introduced. Using the Bohai Bay conditions of regular waves at the same latitude, efficiency ratios are calculated for the irradiated energy between the offshore photovoltaic panels with different dip angles and the onshore ones with the best dip angles. From the calculations, we find that for the panels with different dip angles, variations in the radiation energy under sea waves follow basically the same trend－it decreases with the increase in the average dip angle. A recommended standard of the panel motion amplitude is given as a design criterion useful for estimating hydrodynamic responses in the development of FPV structure.

Previous models for intelligent prediction of rockfill dam deformation, lacking attention to the uneven distribution of deformation time series over multiple measuring points, are limited to low accuracy. This paper develops a rockfill dam deformation prediction model, CTSA-ConvLSTM, to combine a convolutional neural network (CNN), the attention mechanism, and a long short-term memory (LSTM) neural network. This model extracts the temporal and spatial characteristics of deformation and generates different weights for the measurements taken at different instants and different locations, so that it realizes the adaptive learning of global deformation patterns of a rockfill dam. In the case study of the Shuibuya dam, the model is verified against the deformation data from all the measuring points at the maximum dam section. It performs better than Holt-Winters and other conventional time series prediction models, and its prediction accuracy is higher than that of a LSTM-based deformation model developed by the authors. By extracting the spatiotemporal characteristics of monitoring data through deep learning, it improves the accuracy and provides a new idea for improving dam safety monitoring models.

The main objective of this study is to quantify the impacts of peaking flows on riparian habitat fragmentation and macroinvertebrate assemblage variations. Field surveys of the Mudan River were conducted in the autumn of 2014. We sampled its macroinvertebrate assemblages and collected the habitat variables in six selected drawdown pools in its two regulated reaches, and made comparison with the data collected in its natural channels. Redundant analysis (RDA) of the physical and chemical habitat variables is used to summarize total variations in the habitat data and to identify the major environmental gradients in the pools. Results indicate substantial differences in the habitat variables and macroinvertebrate assemblages between the pools and river channels. In the pools, dissolved oxygen and pH were significantly higher than the natural river; higher levels of ammonia nitrogen and total nitrogen were found; and diptera, Hemipteran, Gastropoda were dominant. The abundance, taxa richness, Shannon-Wiener index, and Margalef richness significantly decreased in the pools due to habitat isolation. Redundancy Analysis reveals that the key habitat variables affecting macroinvertebrate assemblages in the pools are dissolved oxygen, pH and total phosphorus.

The seepage monitoring model of rockfill dams is a key factor for quantitative analysis of seepage safety. Most of the traditional models adopt a statistical model or machine learning intelligent algorithm model separately, unable to effectively integrate the advantages of both. This paper presents an innovative integration of statistical models with multiple parallel intelligent algorithm prediction models in the framework of integrated learning, and uses the interpretability of statistical models and the high adaptability of fit of intelligent algorithms to improve the prediction accuracy of this integrated model. First, we fully consider the lag effect of seepage influence factors on the basis of the classical seepage statistical model, and improve the expression for the water level factor and the rainfall factor. Then, based on the integration principle of differential evolution adaptive Metropolis (DREAMZS), several advanced intelligent algorithms and improved statistical models in machine learning are integrated, and optimal weight coefficients are obtained for each model. Case analysis shows that in comparison with the single statistical model or the intelligent algorithm model, our integrated learning fusion model improves prediction accuracy significantly and can integrate effectively the advantages of a statistical model and multiple intelligent models, providing a new modeling method for dam seepage monitoring.

Unsafe behaviors of construction workers are the key factor leading to safety problems in the construction process of water conservancy project. Most construction sites adopt on-site safety inspection, wearable equipment, real-time monitoring, and other methods to identify unsafe behaviors, but such methods are time-consuming, laborious, expensive with low information level, and unfavorable to the timely discovery and early warning of dangerous behaviors. This paper presents a new method of unsafe behavior identification suitable for large scenes of water conservancy projects based on computer vision and deep learning. First, an improved method of YOLOv5 is developed to solve the problem of missing and wrong detection of small objects in construction workers and machinery in large scenes, and a multi-object object detection model of construction workers and machinery is constructed. Then, based on the target detection model, recognition methods are suggested for each of the routine unsafe behaviors, such as being close to the static danger area, dynamic construction machinery, and not wearing safety helmet, etc. Engineering application verifies that our identification method strengthens the means and intensity of construction site control and improves effectively the level of water conservancy engineering construction safety and intelligent control.

This paper develops a new ecological flow forecasting method based on deep learning and a conceptual hydrological model with application to the Jiaojiang River basin in Zhejiang Province to improve the forecast accuracy of ecological flow early warning and the efficiency of ecological operation of water conservancy projects. This method calculates the ecological flow and warning threshold using the hydrological method, and screens model forecast factors through the principal component analysis. The results reveal the check values of most suitable ecological flows are 2.89 m3/s and 1.92 m3/s at the Baizhiao and Shaduan stations, respectively. We use precipitation and evaporation as input factors and the grid search method for optimal parameters searching, and have achieved a 100% qualified rate of the ecological flow warning level forecasts in all the years by using the eXtreme Gradient Boosting (XGBoost) algorithm. Our coupling prediction model based on XGBoost and the Xin’anjiang model can well complete the ecological flow early warning prediction and reservoir ecological flow regulation, laying a basis of decision-making for protection and supervision of water resources in rivers and lakes.

Sediment transport and flow resistance are affected by riverbed morphology. As an important aspect of riverbed evolution, effective predictions of bed form changes are practically significant for river regulation and sediment research. This paper determines the characteristic parameters of riverbed forms based on the law of flow resistance. We find an ‘S’-shaped relationship exists between the characteristic parameters of the flow and bed form through analysis of previous experimental data. By automatic division of the bed forms using a SVM multi-classification method, we obtain a fitting function from the nonlinear and linear fittings of dimensionless characteristic parameters, and determine the criterion of riverbed forms through a derivative analysis of the fitting function of characteristic parameters and the sand wave forms. Finally, this criterion is verified against the previous laboratory experimental data and in-situ measurements in literature, showing the method is feasible and quite accurate in riverbed shape recognition.

The increasing urban flooding threatens urban safety and sustainable development, and further improvement in the accuracy of urban flood risk simulation and assessment are essential to protect people's lives and properties. In this study, a flood simulation model is developed for the Shenzhen River basin based on a Cellular Automata flood simulation model, and inundation depths are simulated for the design rainfall conditions of this basin. Using the Hazard-Vulnerability risk assessment framework, we consider its different rainfall scenarios and assign weights to its different urban flood risks, through a game theory-based combination of weights determined by the analytic hierarchy process and the Criteria Importance Though Intercrieria Correlation methods. Then, its urban flood risks are evaluated and predicted. The results show typical temporal and spatial differences occur over the high-risk areas of urban flooding in the basin. With an increasing rainfall recurrence period, the ratio of the risky area increases, such as from 0.20% of a 2-year rainfall event to 0.82% of a 100-year event; the ratio of the medium-low risk area decreases, such as from 67.4% to 64.7% in terms of the two events. The results of this study can be visualized to show the distribution of flood risks over the basin, helping improve its flood control system and enhance the urban flood prevention capability and resilience of the city.

Effective control of the compaction quality of roller compacted concrete (RCC) is an important prerequisite for ensuring the safety and stable operation of dams. This paper formulates an improved stiffness index of RCC dam material, considering its lag phase angle, and presents a new method for measuring the lag phase angle. Then, we develop a comprehensive evaluation method of RCC compaction quality, using an XGBoost algorithm that couples this index, dam material properties, rolling parameters, and frequency domain compaction indexes. The results of case analysis show that the improved stiffness index can better characterize the compaction quality of dam materials. The comprehensive evaluation model has a relatively high accuracy, with a correlation coefficient of determination of 0.86 and an average relative error of 0.47% in compactness calculation against measurements, thus providing a new effective approach to the compaction quality control of RCC dam construction.

Soil piping is one of the main causes of instability and failure in hydraulic engineering. To date, most of the existing studies on soil piping ignored the shape effect of sand particles; The seepage deformation results of some real projects are inconsistent with the theoretical criterion. In this work, a self-designed soil permeability test device is used to conduct piping tests on a variety of soils that have different gradings and three different particle shapes. Based on the fractal theory and a capillary model, critical hydraulic conditions for soil piping are analyzed at multiple scales, and the testing process is monitored using an acoustic emission system. The results show that sand samples with spherical particle shape have relatively high permeability, prone to piping failure; with the increase of blocking particles, the soil tends to be stable under the action of seepage. The critical hydraulic gradient of soil piping is inversely proportional to the mass fractal dimension; the capillary model based on particle shape and effective pore volume predicts the critical flow rate of piping with a satisfactory accuracy. The evolution patterns of acoustic emission cumulative energy can reflect the characteristics of soil piping development.

Dam-break accidents occur at times at hydropower stations; dam-break waves often impose a significant impact on the downstream structures. In this work, physical model tests are carried out on a flume to record the mechanical characterization of a slope-deck structure, located at the end of the flume that is impacted by the dam-break waves created by a dam-break wave generating system (DWGS). We focus on analysis of the load history and distribution characteristics of uplift pressure on the structure during wave impacting. In time variation, the impact pressure shows five successive stages; the two stages of wave front impact pressure and quasi-constant pressure are examined in this paper. In spatial distribution, these two pressures decrease along the stream from the head to tail of the deck section, while they change little in the cross-flow direction. Based on the test data, a formula for calculating the platform uplift pressure distribution is fitted, and streamwise variations in the peak values of wavefront impact pressure and quasi-constant pressure are obtained.

When a running axial flow pump deviates from its design condition, especially at very low flows, strong vortices in its flow channel will cause severe pressure pulsation, deteriorating the safe and stable operation of the pump unit. In this paper, by examining a high specific speed axial flow pump as a case study, we simulate numerically three-dimensional flows in its whole flow channel for five operating points by using a code equipped with a RNG k-ε turbulence model, and discuss its unsteady flow characteristics. Its external characteristic curves at five steady operating points are verified, and we find they agree well with the design performances. We calculate the unsteady cases at two flows of 0.2Qd and 1.0 Qd. The results show obvious differences occur between the two conditions in the amplitude and frequency of pressure pulsation mixing at the same measuring points. At the flow of 0.2Qd, the maximum peak-to-peak value of pressure pulsation is four times higher than that at 1.0 Qd, indicating highly uneven flow patterns in the channel. In unsteady cases, the dynamic and static interference, usually causing low-frequency pulsation, will greatly enhance the 3fn and 5fn pulsation amplitudes of its principal frequencies. At the measuring points closer to the impeller and guide vanes, these increases are more significant, and the flows are more sensitive to the interference, with the pulsation amplitudes of 3-5 times that of 1.0 Qd; the influence in the runner and guide vane section is stronger. Thus, pressure fluctuation at the lower flow is more severe, and the impeller and guide vane sections are more affected by the static and dynamic interference, so that severe flow separation and more vortices occur in the guide vane channels.

In the context of new-type power systems, their proportion of new energy is increasing gradually. Analysis of the balance of an electric power and energy system faces new challenges because of randomness and fluctuation in new energy generation and its reverse peak regulation. This paper combines the strategic needs of energy transition to develop a new power balance mode for such a power system and demonstrates the key workflow of its power balance. Then, we conduct detailed modeling of its thermal power, hydro-power, pumped storage/storage power supply, and construct a sequential power balance model. Finally, a case study of a real power grid is used to verify this new method. We show that it reflects the characteristics of power sources accurately, gives the conditions of new energy consumption, and supports the study of power system lifting measures, thus significantly raising the level of scientific planning of the power systems.

For the time series prediction issue of dam deformation, a spatiotemporal multi-dimensional input matrix of deformation is derived considering the correlation of deformation at multiple measuring points; an Informer-AD dam deformation prediction model is constructed that integrates multi-dimensional spatiotemporal information based on K-means clustering. We use the K-means clustering to partition rationally the deformation measuring points, then apply a panel data regression model to integrate the analysis of spatiotemporal dimensions and partition results. Finally, we develop an Informer-AD dam deformation prediction model to integrate multi-dimensional spatiotemporal information. This model is used to learn spatial feature sequences and integrate spatial features through a fully connected layer to output predicted dam deformation values. Its application to a concrete gravity dam shows that our prediction method, considering spatiotemporal correlation, can fully explore the relationship of the overall state of dam deformation versus the spatial distribution characteristics of measuring points. It better captures the spatiotemporal characteristics of deformation values and thus improves prediction accuracy, which implies that our model has a high accuracy and satisfactory applicability, useful for engineering application.

To maximize power generation benefit, this study explores reservoir impounding strategies adapted to extreme low-flow conditions in the case of the lower Jinsha River and the Three Gorges cascade reservoirs, and presents a new reservoir impounding scheme by combining the impounding rule of benefit maximization and the water supply requirement of the Three Gorges Reservoir. We also examine alternative reservoir impounding schemes aimed at different water allocation principles that meet the targets of reservoir stages or the control boundaries. Typical extreme low-flow scenarios are simulated numerically for operation and benefit analysis. From the results, we demonstrate a critical condition of natural inflow for full filling of the cascade reservoirs－the natural inflow is above the level at the frequency of 97% from August through early September, and above the level at the frequency of 90% from mid-September through November. Under extreme low-flow conditions, the cascade reservoirs will be able to perfectly fulfill the water supply requirement of the Three Gorges Reservoir if they adopt a certain joint water supply dispatching. In the case of a full filling impossible, their power generation benefit can be maximized by impounding preferentially those in the lower Jinsha River, such as the Baihetan Reservoir first and then the Xiluodu Reservoir as recommended by this study.

When a concentrated leakage passage exists in the body of a dam, the passage will exchange water with its surrounding porous media driven by pressure difference. Such exchange may lead to an increase in dam discharge, but it is often ignored in traditional seepage field analysis. This study develops a sand tank model to simulate the presence of concentrated seepage channels in a dam, and examines the mechanism of water exchange and the water discharges from the concentrated leakage passage and the porous medium. A new formula for calculating the water exchange coefficient is then suggested. Results show that the amount of water exchange is proportional to passage diameter (D), porous medium permeability coefficient (Km), and hydraulic gradient (J). Orthogonal tests are adopted to analyze the influence degree of each factor on the water exchange capacity, revealing D is most influential, followed by J and Km. The results demonstrate the importance of the channel-medium exchange in seepage calculation, and would lay a basis for seepage field analysis of concentrated seepage channels in the dam.