Title: Prediction of Vistula water surface level by applying the new group method of data handling

Authors: Amin Mahdavi-Meymand; Wojciech Sulisz

Addresses: Institute of Hydro-Engineering, Polish Academy of Sciences, Poland ' Institute of Hydro-Engineering, Polish Academy of Sciences, Poland

Abstract: In this study, novel data-driven models were derived and applied to predict water surface level for two stations of the Vistula river. The applied data-driven models comprise of quadratic polynomial group method of data handling (QP-GMDH), nonlinear group method of data handling (N-GMDH), linear regression equation (LRE), and nonlinear regression equation (NRE). These models were trained by particle swarm optimisation (PSO) and teaching-learning-based optimisation (TLBO) meta-heuristic algorithms. The results confirmed the superiority of GMDHs with respect to the regression equations for big and small datasets. According to the results, the N-GMDH derived in the study is about 4.68% more accurate than the standard GMDH. The TLBO algorithm is about 2.06% more accurate than the popular and widely applied PSO approach. The TLBO method is also more stable in reaching global solutions in comparison with the PSO method.

Keywords: Vistula; river; discharge; GMDH; TLBO.

DOI: 10.1504/IJHST.2023.132586

International Journal of Hydrology Science and Technology, 2023 Vol.16 No.2, pp.151 - 170

Received: 13 Sep 2021
Accepted: 17 Mar 2022
Published online: 30 Jul 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article