Comprehensive Framework for Estimation and Management of Reservoir Sedimentation by Applications of Conservation Modelling and Machine Learning Techniques

Abstract

Abstract Comprehensive Framework for Estimation and Management of Reservoir Sedimentation by Applications of Conservation Modelling and Machine Learning Techniques

Muhammad Bilal Idrees Department of Civil and Environmental Engineering The Graduate School of Hanyang University Ph.D. Advisor: Professor Tae-Woong Kim

Keeping in view the climate change brought about by both natural and anthropogenic factors, water conservation is vitally important, now more than ever. Dams play an important role in the storage and conservation of precious water resources. The benefits of dams and storing water in the form of reservoirs behind the dam are multiplex. Nonetheless, some unavoidable disadvantages are also accompanied by dam construction. Sedimentation is the major menace of pooling river water behind the dam body. The sediment mass carried by running river water gets deposited in the reservoir and result in the loss of precious water storage capacity. The sedimentation hiders the sustainable use of reservoirs and also affects dam safety. Also, the useful life of the reservoir is shortened by sediment deposition. The need for the hour is to view and evaluate the reservoir sedimentation problem from a fresh perspective. A thorough and systematic analysis of multiple aspects of reservoir sedimentation from technical and environmental viewpoints is required. An effort has been made in the present research work to fill the research gap in the available literature on reservoir sedimentation. This work encompasses a comprehensive framework for the evaluation and management of reservoir sedimentation. Moreover, the reservoir sediment inflows, sediment deposition, and possible reservoir sediment management measures have been explored with the interdisciplinary application of machine learning approaches and reservoir conservation modelling.

In the first phase of the research, a parameter estimation method for the RESCON model is developed to calculate efficient sediment flushing parameters in a dam reservoir. The parameters including flushing discharge (Qf), frequency (N), duration of flushing (Tf), and reservoir water surface elevation during flushing (Elf) can be calculated by this method. The efficiency of the sediment flushing operation is assessed based upon well-established criteria in the literature of Sediment Balance Ratio (SBR) and Long Term Capacity Ratio (LTCR). The sediment flushing operation is subjected to multiple onsite constraints and several criteria were used for their evaluation. These include Drawdown Ratio (DDR), Sediment Balance Ratio at full Drawdown (SBRd), Top Width Ratio (TWR), and Flushing Width Ratio (FWR). The method has been applied for the demonstration on the Shahpur reservoir in Pakistan. For efficient flushing operation at Shahpur Dam, retrofitting of bottom outlets in the dam body has been recommended to achieve a minimum Qf of 7.5 m3/s. The reservoir water elevation should be drawn to a minimum level of 430 m, and the flushing operation should be performed annually over 15 days. Flushing operation by achieving these parameters would result in 100% sediment balance and 69% of the original reservoir capacity conserved in the long term. The DDR was 0.72, SBRd was 1.58, flushing width ratio was 0.88, TWR was 1.23. The sensitivity analysis has also been performed and it was found that for every increment of 0.5 m3/s in flushing discharge, the amount of sediment flushed increased by 0.020–0.025 million tonnes. Suspended sediment load (SSL) flowing into a reservoir contributes a significant risk factor towards the overall safety of the dam. Owing to the complexity and stochastic nature of sedimentation, accurate prediction of reservoir SSL inflow is still challenging. Moreover, research and application of machine learning (ML) techniques for reservoir sedimentation are still deficient. In the second phase of research, a comprehensive evaluation of six ML models for a reservoir SSL inflow prediction was performed. ML techniques including artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS), radial basis function neural network (RBFNN), support vector machine (SVM), genetic programming (GP), and deep learning (DL) were applied to develop predictive models of daily SSL inflow at Sangju Weir, South Korea. Significant input vectors for each model were selected with streamflow, water temperature, water stage, reservoir outflow for different time lags. Model performances were evaluated using various statistical indices including the coefficient of determination (R2), mean absolute error (MAE), percentage of bias (PBIAS), Willmott index (WI), Nash–Sutcliffe efficiency (NSE), root mean square error (RMSE), and Pearson correlation coefficient (PCC). The best input combinations were found to be unique for each ML model, but all six models performed reasonably well for SSL inflow predictions. ANN model outperformed other models with R2 = 0.821, MAE = 4.244 tons/day, PBIAS = 0.055, WI = 0.891, NSE = 0.991, RMSE = 11.692 tons/day, PCC = 0.826. The models were ranked based on their SSL prediction capabilities as ANN > ANFIS > DL > RBFNN > SVM > GP from best to worst. In the third and final phase of the research, a two-stage complementary modeling approach has been proposed for comprehensive reservoir sediment management. In the first stage, artificial neural network-based models provide real-time reservoir sediment inflow predictions using water inflow, water head, and outflow as input parameters. In the second stage, the parameter estimation method of the RESCON model is applied to hydraulic flushing in a reservoir. This approach was applied to the Sangju Weir and Nakdong River Estuary Barrage (NREB) in South Korea. The annual sediment inflow volumes were 398,144 m3 and 159,298 m3 for the Sangju Weir and NREB sites, respectively. Results from the RESCON model revealed that hydraulic flushing was effective for sediment management at both the Sangju Weir reservoir and the NREB approach channel. Efficient flushing at the Sangju Weir required a flushing discharge of 100 m3/s for 6 days and 40 m of water head. Efficient flushing at the NREB required a flushing discharge of 25 m3/s for 6 days with 1.8 m of water-level drawdown. The findings of this study can provide guidelines for the implementation of sediment flushing operations for reservoir sediment management. Furthermore, the results are expected to be useful for future dam safety and risk assessment, and for achieving sustainability of reservoir operation through comprehensive sediment management. Moreover, the two-stage approach proposed in the third phase of research is useful in attaining the goal of comprehensive reservoir sediment management.