A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Daily Water Level Time Series Prediction Using ECRBM-Based Ensemble Optimized Neural Network Model
Daily water level prediction for rivers is of great significance in flood prevention and enhanced water resources supervision. In order to accurately predict daily water level time series without sufficient data despite the need for large training data sets for neural networks, this paper proposes an innovative daily water level forecasting model, ECRBM-GRU-SSA, which combines the enhanced continuous restricted Boltzmann machine (ECRBM), the gated recurrent neural unit (GRU), and the sparrow search algorithm (SSA). The ECRBM extracts input features and then cooperates with the ensemble strategy to increase the generalization ability of the final model. SSA adjusts model parameters. The contribution of each component to the final prediction result is analyzed using daily water level meteorological data from the Qingxi River. The accuracy of the proposed model is verified by comparing it with basic prediction models like support vector machine (SVM), random forest (RF), and GRU and with improved models such as ECRBM-GRU and GRU-SSA. The indicators RMSE, MAE, R and NSE are improved from 11.5% to 57.3%, 9.3% to 73.6%, 0.5% to 4.6%, and 5.6% to 31.9%, respectively. Therefore, the proposed model provides technical support for staff managing water resources.
Daily Water Level Time Series Prediction Using ECRBM-Based Ensemble Optimized Neural Network Model
Daily water level prediction for rivers is of great significance in flood prevention and enhanced water resources supervision. In order to accurately predict daily water level time series without sufficient data despite the need for large training data sets for neural networks, this paper proposes an innovative daily water level forecasting model, ECRBM-GRU-SSA, which combines the enhanced continuous restricted Boltzmann machine (ECRBM), the gated recurrent neural unit (GRU), and the sparrow search algorithm (SSA). The ECRBM extracts input features and then cooperates with the ensemble strategy to increase the generalization ability of the final model. SSA adjusts model parameters. The contribution of each component to the final prediction result is analyzed using daily water level meteorological data from the Qingxi River. The accuracy of the proposed model is verified by comparing it with basic prediction models like support vector machine (SVM), random forest (RF), and GRU and with improved models such as ECRBM-GRU and GRU-SSA. The indicators RMSE, MAE, R and NSE are improved from 11.5% to 57.3%, 9.3% to 73.6%, 0.5% to 4.6%, and 5.6% to 31.9%, respectively. Therefore, the proposed model provides technical support for staff managing water resources.
Daily Water Level Time Series Prediction Using ECRBM-Based Ensemble Optimized Neural Network Model
J. Hydrol. Eng.
Fu, Yi (author) / Zhou, Xinzhi (author) / Li, Bo (author) / Zhang, Yuexin (author)
2023-01-01
Article (Journal)
Electronic Resource
English
Deep Learning Neural Network for Time Series Water Level Forecasting
| Springer Verlag | 2021
Nonlinear ensemble prediction of chaotic daily rainfall
| British Library Online Contents | 2010