A Cascaded Adaptive Network-Based Fuzzy Inference System for Hydropower Forecasting

Area of Science:

• Environmental Science
• Climate Modeling
• Renewable Energy Systems

Background:

• Hydropower is a critical renewable energy source globally.
• Accurate forecasting of hydropower generation is essential for energy management and grid stability.
• Climate change poses significant challenges to consistent water availability for hydropower projects.

Purpose of the Study:

• To assess the impact of climate change on the Samanalawewa Reservoir Hydropower Project's future power generation.
• To evaluate the efficacy of a novel Cascaded ANFIS algorithm for predicting hydropower output.
• To compare the performance of Cascaded ANFIS against state-of-the-art regression models.

Main Methods:

• Utilized rainfall data from selected weather stations within the catchment.
• Generated future rainfall scenarios using bias-corrected Global Climate Models (RCP4.5 and RCP8.5).
• Employed a Cascaded ANFIS architecture for regression analysis, comparing it with Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) models.

Main Results:

• The Cascaded ANFIS algorithm achieved a minimum prediction error of 1.01 for power generation.
• GRU, the second-best model, recorded an error rate of 6.5.
• The algorithm demonstrated high accuracy in forecasting power generation variations linked to rainfall, even under different climate scenarios.

Conclusions:

• The Cascaded ANFIS algorithm is highly effective for predicting hydropower generation, showing superior performance compared to other advanced regression models.
• This research provides a reliable method for forecasting hydropower output under changing climate conditions.
• The findings support informed decision-making and planning for sustainable hydropower development.