Reservoir water level forecasting using wavelet support vector regression (WSVR) based on teaching learning-based optimization algorithm (TLBO)

Mohammad Mahdi Malekpour

doi:10.1007/s00500-022-07296-1

Return

Reservoir water level forecasting using wavelet support vector regression (WSVR) based on teaching learning-based optimization algorithm (TLBO)

2022-07-05

PRE

Mohammad Mahdi Malekpour *

Hossein Malekpoor

DOI：10.1007/s00500-022-07296-1

Original

Original request for help

Save

Abstract

En 中文

As the importance of freshwater preservation, a new hybrid approach is applied in this study. The wavelet support vector regression based on the teaching learning-based algorithm (WSVR-TLBO) is used as the proposed approach for the water level prediction of the Zayanderud dam in Iran. This model is promoted by the wavelet transform and the optimization algorithm which have prompted the error reduction and the accuracy promotion of the SVR model. In the hybrid model, the correlation coefficient (R) and mean square error (MSE) are improved by 12% and 94%, respectively, rather than SVR. The four error criteria are employed for evaluation, and their results are ameliorated in the use of the WSVR-TLBO model. Besides the SVR model, the feed-forward neural network (FFNN), autoregressive integrated moving average (ARIMA), and generalized regression neural network (GRNN) models are also applied to forecast the reservoir water level. The comparison of these models with the hybrid one is performed, and the results show the superiority of the hybrid model. The current approach's error criteria (MSE) are decreased by 67%, 94%, 92%, 92%, and 90% rather than the WSVR, SVR, ARIMA, GRNN, and FFNN models, respectively. All the error criteria reveal that the hybrid approach of this study significantly forecasted the reservoir water level with high accuracy and is outperformed by other compared models.

Keywords:

Forecasting

Hybrid model

TLBO algorithm

Wavelet transform

Reservoir water level

Zayanderud dam