A Comparative Study on Prediction of Monthly Streamflow Using Hybrid ANFIS-PSO Approaches

Monthly prediction of streamflow is a fundamental and complex hydrological phenomenon. Accurate streamflow prediction helps in water resources planning, design, and management, particularly for hydropower production, irrigation, protection of dams and flood risk management. Hence, in this paper, a h...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:KSCE Journal of Civil Engineering Jg. 25; H. 10; S. 4032 - 4043
Hauptverfasser: Samanataray, Sandeep, Sahoo, Abinash
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Seoul Korean Society of Civil Engineers 01.10.2021
Springer Nature B.V
대한토목학회
Schlagworte:
Accuracy
Adaptive systems
Algorithms
Artificial neural networks
Civil Engineering
Comparative analysis
Comparative studies
Engineering
Environmental risk
Evolutionary algorithms
Flood irrigation
Flood management
Fuzzy logic
Geotechnical Engineering & Applied Earth Sciences
Hydroelectric power
Industrial Pollution Prevention
Monthly
Neural networks
Particle swarm optimization
Predictions
Risk management
River basins
Root-mean-square errors
Stream discharge
Stream flow
Water resources
Water Resources and Hydrologic Engineering
토목공학
Barak River
ANFIS-PSO
Neural network
Streamflow
Prediction
ISSN:1226-7988, 1976-3808
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Monthly prediction of streamflow is a fundamental and complex hydrological phenomenon. Accurate streamflow prediction helps in water resources planning, design, and management, particularly for hydropower production, irrigation, protection of dams and flood risk management. Hence, in this paper, a hybrid robust model integrating adaptive neuro-fuzzy inference system (ANFIS) with particle swarm optimization (PSO) algorithm was developed for monthly streamflow prediction of Barak River basin, India. Multiple factors like Precipitation (P t ), temperature (T t ), humidity (H t ), Infiltration loss (I t ), are considered as the inputs for determining the streamflow. For validating the model performance, 70% of data (1980–2007) were used for training them and 30% of data (2008–2019) were used for testing them. A comparison is made between results of developed hybrid model with simple artificial neural network (ANN) and ANFIS models for assessing accuracy and efficiency of model performances. Obtained results of proposed models were evaluated based on four assessment indices including by Root Mean Square Error (RMSE), Mean Absolute Error (MAE), determination coefficient (R 2 ) and Nash-Sutcliffe Coefficient (E NS ). Based on comparison of results, it was concluded that robust ANFIS-PSO model with RMSE = 5.887, MAE = 4.978, R 2 = 0.9668, and E NS = 0.961 demonstrated best performance with more reliability and accuracy in comparison to ANFIS and ANN models. Findings of this research proved that hybrid ANFIS with an evolutionary optimization algorithm is a reliable modelling approach for monthly streamflow prediction.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1226-7988
1976-3808
DOI:10.1007/s12205-021-2223-y