Determination of Deep Learning Model and Optimum Length of Training Data in the River with Large Fluctuations in Flow Rates

Abstract

Recently, developing countries have steadily been pushing for the construction of stream-oriented smart cities, breaking away from the existing old-town-centered development in the past. Due to the accelerating effects of climate change along with such urbanization, it is imperative for urban rivers to establish a flood warning system that can predict the amount of high flow rates of accuracy in engineering, compared to using the existing Computational Fluid Dynamics (CFD) models for disaster prevention. In this study, in the case of streams where missing data existed or only small observations were obtained, the variation in flow rates could be predicted with only the appropriate deep learning models, using only limited time series flow data. In addition, the selected deep learning model allowed the minimum number of input learning data to be determined. In this study, the time series flow rates were predicted by applying the deep learning models to the Han River, which is a highly urbanized stream that flows through the capital of Korea, Seoul and has a large seasonal variation in the flow rate. The deep learning models used are Convolution Neural Network (CNN), Simple Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Bidirectional LSTM (Bi-LSTM) and Gated Recurrent Unit (GRU). Sequence lengths for time series runoff data were determined first to assess the accuracy and applicability of the deep learning models. By analyzing the forecast results of the outflow data of the Han River, sequence length for 14 days was appropriate in terms of the predicted accuracy of the model. In addition, the GRU model is effective for deep learning models that use time series data of the region with large fluctuations in flow rates, such as the Han River. Furthermore, through this study, it was possible to propose the minimum number of training data that could provide flood warning system with an effective flood forecasting system although the number of input data such as flow rates secured in new towns developed around rivers was insufficient.

Development of Deep Learning Models to Improve the Accuracy of Water Levels Time Series Prediction through Multivariate Hydrological Data

“…Among the DNN models that can be effectively used for time-series data analysis, in this study, Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) were selected and used as Recurrent Neural Network (RNN) models that can be appropriately applied to past hydrological time-series data [4].…”

“…In the past, flow rates and water levels were calculated in the event of a flood, using the physical numerical model that used the Computational Fluid Dynamics (CFD) method, but there were limitations in obtaining sufficiently reliable results in terms of time, cost, and accuracy of the prediction model [4]. Therefore, in the field of water resource engineering, as a way to replace the existing physical models and improve the prediction accuracy of hydrological quantities, researchers have developed data-driven models that can predict hydrological quantities only through the analysis of input data.…”

“…Accordingly, studies on data-driven models based on the Fourth Industrial Revolution technology that can overcome and replace the limitations of existing physical models are currently being actively studied [4]. A typical data-driven model is a Deep Neural Network (DNN) model, and in contrast to the calculations of numerical solution according to the initial and boundary conditions of the governing equation, accurate prediction results can be obtained by repeatedly training input value data and calculating parameters of the data-driven model.…”

“…Especially, DNN was widely used to predict groundwater levels [20], water levels in the stream [21][22][23], wetlands [24,25], and reservoirs [26][27][28]. However, most of these studies on water-level predictions correspond to cases where the water-level change according to temporal changes is mild, and the accuracy of the prediction results could not be guaranteed if the water-level change changes rapidly [4]. Therefore, the studies on predicting water levels with high temporal variability in the rivers were recognized as limitations of the DNN models and were rarely conducted, as compared to other research areas.…”

“…It is essential to predict the exact amount of hydrological quantities in the river due to torrential rains or typhoons caused by climate change and urbanization, but it was difficult to trust the prediction results of DNN models under high-water-level or high-flow-rate conditions. Nevertheless, there was a study to predict river flow rates with high accuracy by providing guidelines for input training data [4]. In addition, real-time groundwater-level prediction [29], extreme water-level prediction by highest tide [30,31], and real-time urban-river-flood-level prediction, using various DNN models, have been effectively performed recently [32].…”