RESERVOIR OPERATIONS USING HYBRID MODELLING - ANN AND FNN

Abstract

In this study, an attempt has been made to study reservoir operations using different technologies to check for their efficiency in predicting the water supply release and comparing it with the demands to be met. A hybrid modelling comprising of ANN and FNN has been used. ANN has the ability to predict the output data by assigning weights to the input variables based on its learning ability while FNN has an advantage of making decisions based on the logical inferences of fuzzy rules. The system considered for study purposes in this project is multipurpose reservoir dam across Sutlej river- Bhakra reservoir also known as Govind Sagar Reservoir situated in Una and Bilaspur district of Himachal Pradesh. Combining these two model for reservoir operations of Bhakra reservoir will improve the efficiency of operations of reservoir during critical time period. From the results it was observed that the improvement in the performance of reservoir operations were significant emphasizing on the use of AI methods such as ANN and FNN. Storage capacity, inflow into the reservoir and Evaporation losses were considered as independent parameters for the modelling the operations of ANN and FNN. These input parameters are tested for randomness and any correlations using the regression analysis method. Linear regression method for prediction of release is also effective but is not suggested because of numerous computations involved, this may result in confusion and chaos. Hence, the hybrid model of ANN and FNN is regarded as the most suited one for reservoir operations. Along with this an attempt has been made to predict the release and compare the same graphically using programming language such as Python. The algorithms of ANN, Random Forest and XGBosst has been used as these operates on the principle of Decision Trees and hence are considered effective for prediction and comparison purpose.