Publication:
Application of augmented bat algorithm with artificial neural network in forecasting river inflow of hydroelectric reservoir stations in Malaysia

dc.contributor.affiliationen_US
dc.contributor.authorJoe Wee Wei, Mr.en_US
dc.date.accessioned2023-05-03T13:43:44Z
dc.date.available2023-05-03T13:43:44Z
dc.date.issued2022-02
dc.description.abstractThe river is one of our main sources of water which plays a huge role in our daily lives. Hydrologists depend heavily on river streamflow (SF) forecasting to monitor and control flood management and water demand for the inhabitants. Unfortunately, one of the main issues faced by the hydrologists is the effect of the changes that occur to the river. Climate change and rainfall greatly affect the river SF, which would then cause issues to nearby reservoir stations that rely on the inflow of the river. Despite there is only a limited number of simulation systems, existing methods failed to efficiently foresee the SF data and the methods are not cost-effective and takes a long time to carry out. Artificial intelligence (AI) was implemented to build forecasting models. However, standalone models, such as Artificial Neural Network (ANN), typically underperform during local convergence and resulted in slow-speed convergence. Thus, Bat Algorithm (BA) was used to enhance the efficiency of ANN in forecasting upstream river SF, as BA is capable of switching from the “explore to exploit” function which could increase the rate of convergence at the initial stage and deliver a quick result for a majority of a classification problem. This research aims to develop and utilize ANN Standalone and produce a Hybrid BA-ANN model to forecast future river streamflow (SF) at the upstream of five selected hydroelectric reservoir stations in Malaysia. Both standalone and hybrid models were developed to identify the most optimum parameter to be used for river SF forecasting. Then, simulations were carried out to identify the most effective and accurate model for future river SF forecasting. Different parameters and scenario cases also played a significant part in forecasting future river SF Sensitivity Analysis simulation. The performance indicators used were RMSE, R, R2, and MAE. The statistical tests analysis from the preliminary results demonstrated that Hybrid BAANN was more superior to forecast the SF at all five selected study areas, having the RMSE value of 0.085 m3/s for training and 0.113 m3/s for testing, whereas 0.093 m3/s and 0.121 m3/s were obtained from the training and testing of ANN standalone, respectively. R values for both training and testing are obtained at 0.980 and 0.974 respectively, while ANN standalone produced R value of 0.973 for training and 0.62 for testing. As for R2, it is at 0.983 for the training data set and R2 of 0.948 for the testing data set of the hybrid model, compared to 0.955 and 0.932 for traing and testing data set of ANN standalone. The MAE values of the hybrid model are almost zero, at 0.004 m3/s and 0.006 m3/s for training and testing, respectively. Uncertainty Analyses such as Taylor Diagram, Violin Plot, Relative Error, and Scatter Plot were applied to further validate the results. The Hybrid BA-ANN model proved to be versatile and robust when being applied to other study areas in Malaysia. Once optimum results had been obtained at the preliminary station; Kenyir, the simulations had been applied to other four stations, namely Bersia station, Chenderoh station, Kenering station, and Temenggor station, which also managed to produce promising results. To conclude, the results reported here could assist in enhancing ANN for future river SF forecasting by using BA as it was proven to enhance the performance of ANN Standalone which was utilized as an engineering tool for river SF forecasting.en_US
dc.identifier.urihttps://irepository.uniten.edu.my/handle/123456789/19650
dc.language.isoenen_US
dc.subjectAugmented bat algorithmen_US
dc.titleApplication of augmented bat algorithm with artificial neural network in forecasting river inflow of hydroelectric reservoir stations in Malaysiaen_US
dc.typeResource Types::text::Thesisen_US
dspace.entity.typePublication
Files