Implementation of supervised intelligence committee machine method for monthly water level prediction
- PDF / 6,880,740 Bytes
- 14 Pages / 595.276 x 790.866 pts Page_size
- 78 Downloads / 168 Views
ORIGINAL PAPER
Implementation of supervised intelligence committee machine method for monthly water level prediction Mohammad Mahdi Malekpour 1
&
Mahmoud Mohammad Rezapour Tabari 2
Received: 4 April 2020 / Accepted: 15 September 2020 # Saudi Society for Geosciences 2020
Abstract The correct prediction of reservoirs water level variation is one of the important issues for designing, operation of dams, and water supply management. In this study, based on four soft models which are the support vector regression (SVR), adaptive neurofuzzy inference system (ANFIS), artificial neural network (ANN), radial basis function neural network (RBFNN), and combinatory use of their results as input to one of these four models, a new structure is proposed. It is named supervised intelligence committee machine (SICM) for monthly reservoir water level prediction of the Karaj Amirkabir dam. Evaluation of the above models is performed by nine error criteria and eventually the best model among them is selected by the vikor decision-making method. The supervised support vector regression (SICM-SVR) is shown high accurate in monthly prediction rather than SVR model with increasing the Nash-Sutcliffe efficiency (NS) from 0.58 to 0.81 (over 39% increase) and decreasing the mean square error (MSE) from 117.8 to 55.78 m2 (over 52% decrease). According to the vikor analysis among all soft and hybrid models, the SICM-ANN is selected as the best model with NS and MSE equal to 0.94 and 12.85 m2, respectively. Generally, the proposed method results show that all supervised (hybrid) models have higher performance than soft ones and can be effectively applied to reduce the predicted error of water level. Keywords Prediction . Reservoir water level . Karaj dam . Supervised intelligence committee machine . Soft models
Introduction In recent decades, the water crisis and shortage of freshwater resources have been led to many problems in water management and planning, especially during dry periods. For this purpose, the surface water reservoirs of dams play a key role in supplying downstream water needs as one of the essential infrastructures. Due to the dependence of suitable storage of water volume in reservoirs on hydrological parameters, the development of prediction models can be Responsible Editor: Broder J. Merkel * Mohammad Mahdi Malekpour [email protected] Mahmoud Mohammad Rezapour Tabari [email protected] 1
Water and Hydraulic Structures, Shahrekord University, Shahrekord, Iran
2
Department of Engineering, University of Mazandaran, Mazandaran, Iran
significantly regarded for the accurate study of water level behavior (Sammen et al. 2017; Hipni et al. 2013). For this purpose, the use of artificial intelligence (AI) models have been considerably developed in recent years to predict the hydrological parameters. These models do not need to accurately describe physical parameters and they train simply the systems based on the relations between inputs and outputs (Das et al. 2016). These models are known as soft models (SM) which inclu
Data Loading...
