Comparison of Machine Learning Algorithms for Discharge Prediction of Multipurpose Dam

For effective water management in the downstream area of a dam, it is necessary to estimate the amount of discharge from the dam to quantify the flow downstream of the dam. In this study, a machine learning model was constructed to predict the amount of discharge from Soyang River Dam using precipitation and dam inflow/discharge data from 1980 to 2020. Decision tree, multilayer perceptron, random forest, gradient boosting, RNN-LSTM, and CNN-LSTM were used as algorithms. The RNN-LSTM model achieved a Nash–Sutcliffe efficiency (NSE) of 0.796, root-mean-squared error (RMSE) of 48.996 m3/s, mean absolute error (MAE) of 10.024 m3/s, R of 0.898, and R2 of 0.807, showing the best results in dam discharge prediction. The prediction of dam discharge using machine learning algorithms showed that it is possible to predict the amount of discharge, addressing limitations of physical models, such as the difficulty in applying human activity schedules and the need for various input data.

Monitoring Inflow Dynamics in a Multipurpose Dam Based on Travel-Time Principle

Understanding inflow dynamics in a dam lake forms the basis for optimal dam operation and management practices. However, methods pertaining to adequately determining negative inflows and addressing them, as well as quantifying uncertainties in dam inflow, have been scarcely investigated. In this study, the inflow was observed using two pairs of fluvial acoustic tomography (FAT) systems placed diagonally in a dam lake, forming a crossed-shaped pattern. The “travel-time” principle is the primary approach for measuring the inflow by FAT. The novelty of this study is in discussing the inflow characteristics within a slow water-flow environment monitored by FAT. Based on the reciprocal sound transmission, we upgraded an equation to estimate the flow direction; this newly proposed generalized equation can be used in a fluctuating flow environment. We also discussed the sound propagation characteristics for slow flow velocities. Finally, we demonstrated that a small inaccuracy in the acoustic signal, even by a sub-millisecond, can cause significant errors in measurements. One of the novel findings of this study is the detection of internal waves using the improved flow direction equation and acoustic travel-time records. Overall, this study presents a promising approach for inflow measurements under extremely slow flow conditions.

Improving the Accuracy of Dam Inflow Predictions Using a Long Short-Term Memory Network Coupled with Wavelet Transform and Predictor Selection

Accurate and reliable dam inflow prediction models are essential for effective reservoir operation and management. This study presents a data-driven model that couples a long short-term memory (LSTM) network with robust input predictor selection, input reconstruction by wavelet transformation, and efficient hyper-parameter optimization by K-fold cross-validation and the random search. First, a robust analysis using a “correlation threshold” for partial autocorrelation and cross-correlation functions is proposed, and only variables greater than this threshold are selected as input predictors and their time lags. This analysis indicates that a model trained on a threshold of 0.4 returns the highest Nash–Sutcliffe efficiency value; as a result, six principal inputs are selected. Second, using additional subseries reconstructed by the wavelet transform improves predictability, particularly for flow peak. The peak error values of LSTM with the transform are approximately one-half to one-quarter the size of those without the transform. Third, for a K of 5 as determined by the Silhouette coefficients and the distortion score, the wavelet-transformed LSTMs require a larger number of hidden units, epochs, dropout, and batch size. This complex configuration is needed because the amount of inputs used by these LSTMs is five times greater than that of other models. Last, an evaluation of accuracy performance reveals that the model proposed in this study, called SWLSTM, provides superior predictions of the daily inflow of the Hwacheon dam in South Korea compared with three other LSTM models by 84%, 78%, and 65%. These results strengthen the potential of data-driven models for efficient and effective reservoir inflow predictions, and should help policy-makers and operators better manage their reservoir operations.

Applicability of Lindley Distribution for Hydroelectric Dam Operations in the Dry Season

The long-term low-flow data are necessary for efficient planning of water resources and for estimating accurate quantiles via runoff data analysis at point. However, the short recording time period, low confidence, inconsistent distribution model, and parameter estimation method, make it difficult to estimate a proper low-flow quantile for each return period. In this study, the Lindley distribution model, which is a mix of the exponential and the gamma distribution models and has been verified as efficient by previous studies, was used to analyze the low-flow frequency using dam inflow data. We studied its applicability via comparison with statistics of observed data and other models already used for low-flow frequency analysis. For this, we carried out a performance analysis through a low-flow frequency analysis of inflow data from the hydroelectric dam and the reappearance capacity assessment of observed data at the Han river watershed. As a result, the hydrological applicability of the Lindley distribution model and its relative qualitative and quantitative excellence compared to the existing model were verified.

Development and Evaluation of the Combined Machine Learning Models for the Prediction of Dam Inflow

Predicting dam inflow is necessary for effective water management. This study created machine learning algorithms to predict the amount of inflow into the Soyang River Dam in South Korea, using weather and dam inflow data for 40 years. A total of six algorithms were used, as follows: decision tree (DT), multilayer perceptron (MLP), random forest (RF), gradient boosting (GB), recurrent neural network–long short-term memory (RNN–LSTM), and convolutional neural network–LSTM (CNN–LSTM). Among these models, the multilayer perceptron model showed the best results in predicting dam inflow, with the Nash–Sutcliffe efficiency (NSE) value of 0.812, root mean squared errors (RMSE) of 77.218 m3/s, mean absolute error (MAE) of 29.034 m3/s, correlation coefficient (R) of 0.924, and determination coefficient (R2) of 0.817. However, when the amount of dam inflow is below 100 m3/s, the ensemble models (random forest and gradient boosting models) performed better than MLP for the prediction of dam inflow. Therefore, two combined machine learning (CombML) models (RF_MLP and GB_MLP) were developed for the prediction of the dam inflow using the ensemble methods (RF and GB) at precipitation below 16 mm, and the MLP at precipitation above 16 mm. The precipitation of 16 mm is the average daily precipitation at the inflow of 100 m3/s or more. The results show the accuracy verification results of NSE 0.857, RMSE 68.417 m3/s, MAE 18.063 m3/s, R 0.927, and R2 0.859 in RF_MLP, and NSE 0.829, RMSE 73.918 m3/s, MAE 18.093 m3/s, R 0.912, and R2 0.831 in GB_MLP, which infers that the combination of the models predicts the dam inflow the most accurately. CombML algorithms showed that it is possible to predict inflow through inflow learning, considering flow characteristics such as flow regimes, by combining several machine learning algorithms.

Linking Hydraulic Modeling with a Machine Learning Approach for Extreme Flood Prediction and Response

An emergency action plan (EAP) for reservoirs and urban areas downstream of dams can alleviate damage caused by extreme flooding. An EAP is a disaster action plan that can designate evacuation paths for vulnerable districts. Generally, calculation of dam-break discharge in accordance with dam inflow conditions, calculation of maximum water surface elevation as per hydraulic channel routing, and flood map generation using topographical data are prepared for the purposes of creating an EAP. However, rainfall and flood patterns exhibited in the context of climate change can be extremely diverse. In order to prepare an efficient flood response, techniques should be considered that are capable of generating flood maps promptly while taking dam inflow conditions into account. Therefore, this study aims to propose methodology that is capable of generating flood maps rapidly for any dam inflow conditions. The proposed methodology was performed by linking a dynamic numerical analysis model (DAMBRK) with a random forest regression technique. The previous standard method of drawing flood maps often requires a significant amount of time depending on accuracy and personnel availability; however, the technique proposed here is capable of generating a flood map within one minute. Through use of this methodology, the time taken to prepare flood maps in large-scale water-disaster situations can be reduced. Moreover, methodology for estimating flood risk via use of flood mapping has been proposed. This study would provide assistance in establishing disaster countermeasures that take various flood scenarios into account by promptly providing flood inundation information to disaster-related agencies.

Evaluation of Water Quality Interaction by Dam and Weir Operation Using SWAT in the Nakdong River Basin of South Korea

The purpose of this study was to evaluate the streamflow and water quality (SS, T-N, and T-P) interaction of the Nakdong river basin (23,609.3 km2) by simulating dam and weir operation scenarios using the Soil and Water Assessment Tool (SWAT). The operation scenarios tested were dam control (Scenario 1), dam control and weir gate control (Scenario 2), dam control and sequential release of the weirs with a one-month interval between each weir (Scenario 3), dam control and weir gate full open (Scenario 4), dam control and weir gate sequential full open (Scenario 5), weir gate control (Scenario 6), weir gate full open (Scenario 7), and weir gate sequential full open (Scenario 8). Before evaluation, the SWAT was calibrated and validated using 13 years (2005–2017) of daily multi-purpose dam inflow data from five locations ((Andong Dam (ADD), Imha Dam (IHD), Hapcheon Dam (HCD), Namkang Dam (NKD), and Milyang Dam (MYD))multi-function weir inflow data from seven locations (Sangju Weir (SJW), Gumi Weir (GMW), Chilgok Weir (CGW), Gangjeong-Goryeong Weir (GJW), Dalseong Weir (DSW), Hapcheon-Changnyeong Weir (HCW), and Changnyeong-Haman Weir (HAW)), and monthly water quality monitoring data from six locations (Andong-4 (AD-4), Sangju (SJ-2), Waegwan (WG), Hapcheon (HC), Namkang-4 (NK-4), and Mulgeum (MG). For the dam inflows and dam storage, the Nash-Sutcliffe efficiency (NSE) was 0.59~0.78, and the coefficient of determination (R2) was 0.71~0.90. For water quality, the R2 values of SS, T-N, and T-P were 0.58~0.83, 0.53~0.68, and 0.56~0.79, respectively. For the eight dam and weir release scenarios suggested by the Ministry of Environment, Scenarios 4 and 8 exhibited water quality improvement effects compared to the observed data.

Mean areal rainfall improvement using radar rainfall estimation technique by considering geographic character for dam operation

<p>Efficiently dam operation is necessary to secure water resources and to respond to floods. For the dam operation, the amount of dam inflow should be accurately calculate. Rainfall information is important for the amount of dam inflow estimation and prediction therefore rainfall should be observed accurately. However, it is difficult to observe the rainfall due to poor density of rain gauges because of the dam is located in the mountainous region. Moreover, ground raingauges are limitted to localized heavy rainfall, which is increasing in frequency due to climate changes. The advantage of radar is that it can obtain high-resolution grid rainfall data because radar can observe the spatial distribution of rainfall. The radar rainfall are less accurate than ground gauge data. For the accuracy improvement of radar rainfall, many adjustment methods using ground gauges, have been suggested. For dam basin, because the density of ground gauge is low, there are limitations when apply the bias adjustment methods. Especially, the localized heavy rainfall occurred in the mountainous area depending on the topography. In this study, we will develop a radar rainfall adjustment method considering the orographic effect. The method considers the elevation to obtain kriged rainfall and apply conditional merging skill for the accuracy improvement of the radar rainfall. Based on this method, we are going to estimate the mean areal precipitation for hydropower dam basin. And, we will compare and evaluate the results of various adjustment methods in term of mean areal precipitation and dam inflow.</p><p>This work was supported by KOREA HYDRO & NUCLEAR POWER CO., LTD (No. 2018-Tech-20)</p><div> </div><div> </div>

Research on dam inflow analysis based on radar rainfall data

<p>Estimation of dam inflow using rainfall needs for efficient and timely operation of dam. Accuracy rainfall data is important to estimate dam inflow. Currently, rainfall pattern has volatile temporal and spatial distribution. Dam inflow based on rainfall gauged data is inadequate for operating hydroelectric dam. Radar rainfall has been used as an alternative because radar data provides spatially distributed rainfall. In this study, we estimated inflow discharge for hydroelectric dam using both radar and rain gauged data to find a case to improve the accuracy. Hydrological modeling have been adopted to estimate inflow and based on rainfall data collected from 2018 to 2019.</p><p>This work was supported by KOREA HYDRO & NUCLEAR POWER CO., LTD(No. 2018-Tech-20)</p>