Deep Learning Models for Future Hydropower Production Prediction: a Case Study of the Koka Hydroelectric Power Plant, Ethiopia
No Thumbnail Available
Date
2024-06
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Hydropower, a clean and renewable energy source driven by the water cycle, plays a crucial role in many countries. Predicting future hydropower production is vital for strategic decision-making and optimizing energy resource utilization. This study evaluated three deep learning models (Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), and Gate Recurrent Unit (GRU)) for daily and weekly hydropower energy production future prediction at the Koka Dam in Ethiopia’s Awash basin. The models aimed to forecast not only daily and weekly production but also the next year of generation. Daily and weekly hydropower
production, precipitation, minimum temperature, maximum temperature, relative humidity, maximum wind speed, minimum wind speed, wind direction, and all
sky-surface short wave data from September 2010 to November 2023 were used.
After preprocessing, the data was split into training and testing sets for model training and evaluation respectively. Performance metrics like R-squared, MAE,
MSE, and RMSE were calculated for each model. The GRU model emerged as the best performer, achieving an R-squared value of 0.9920, MAE of 8.6121, MSE
of 143.5549, and RMSE of 11.9814 for daily hydropower energy generation prediction, and R-squared of 0.9960, MAE of 5.9925, MSE of 66.2187, and RMSE of
8.1375 for weekly hydropower energy generation prediction. This superior model was then employed to predict the Koka Dam’s daily and weekly hydropower energy production for the next 1 year. The study identified the right hydro-power production prediction model for future potential. Furthermore, this outcome can
help to maximize the use of medium hydro-power resources, contributing to the region’s energy security and sustainable development.
Description
Keywords
Hydro-Power, GRU, LSTM, RNN, Metrics, Prediction