Implementation of machine learning algorithms for streamflow prediction of Dokan dam

• 主要作者: Sarmad Dashti Latif, Mr.
• 格式: text::Thesis
• 语言: English
• 出版: 2023
• 主题: Implementation of machine learning algorithms for streamflow prediction of Dokan dam

Dam and reservoirs play an important role in the control and management of water resources; they have benefited human societies in many ways, allowing for improved human health, expanded food production, water supply for domestic and industrial use, economic growth, irrigation, hydro-power generation, and flood control. Correct inflow forecast is an essential non-engineering measure to confirm flood-control protection and raise water supply efficiency. In addition, accurate inflow prediction can offer reservoir planning and management guidance since inflow is the major input into reservoirs. This study aims at comparing the application of deep learning algorithms and conventional machine learning algorithms for predicting reservoir inflow. Daily inflow and rainfall time-series data have been collected as two hydrological parameters to forecast reservoir inflow using the developed deep learning long-short term memory (LSTM) model and conventional machine learning models, namely support vector machine (SVM), random forest (RF), and boosted regression tree (BRT). Two scenarios with different time-lags (daily, weekly, and monthly) and input combinations have been selected for this study. The first scenario is to apply inflow data as model input for predicting inflow (output). The second scenario is to apply inflow and rainfall data as model inputs to predict inflow (output). The input combinations are selected based on the auto-correlation function (ACF). Dokan dam in Iraq was selected as the case study for this research. The data is collected from the ministry of agriculture and water resources, Kurdistan regional government, Iraq. For the purpose of generalization, the proposed models have been applied to the Warragamba dam in Sydney, Australia, as a different climate condition. Seven statistical indices have been selected to evaluate the performance of the proposed models. The selected statistical indices are root mean square error (RMSE), mean absolute error (MAE), mean square error (MSE), correlation coefficient (R), and coefficient of determination (R2), Nash Sutcliffe Model Efficiency Coefficient (NSE), and the RMSE-observations standard deviation ratio (RSR). The results showed that the first scenario is the most accurate one, and LSTM model outperformed other proposed models with a significant difference with R2=0.99, NSE=0.98, and RSR=0.14. In the second study location, the results from LSTM showed a huge difference in accuracy compared to the other models with R2=0.99, NSE=0.99, and RSR=0.05. The findings indicate that the proposed LSTM model could be generalized and applied to other dams in different locations in order to successfully forecast reservoir inflow.