Hydro-power generation forecast in South Africa based on Machine Learning (ML) models

Selaki Ivy Ramarope, Olawale Samuel Fatoba, Tien Chien Jen

Research output: Contribution to journalArticlepeer-review


With the advancement of technology and the ever-growing need for electronics, electricity has become an indispensable aspect of modern life. Developing or underdeveloped nations must overcome a number of obstacles to balance the supply and demand for electricity. The difference between supply and demand for electricity has a significant impact on economic growth. Energy forecasting will be crucial in helping policymakers recognize unanticipated changes in the demand for electricity under certain circumstances in a timely manner. Energy constitutes one of the most critical components in a country's growth. To achieve this, machine learning (ML) model was created for hydropower plant electricity generation predictions. In this study, the forecasting model examined the meteorological variables in South Africa. The data included how the South African Weather Service categorized different years from 2001 to 2019 in the southern hemisphere based on climatological and social factors. The simulation allows for an experimental paradigm for the established model to explore the behaviour and performance of the components effective for prospective saving variables related to the design and applications of forecasting hydropower generation in South Africa. In addition, a variety of time-series-based models was used and supervised machine learning techniques to predict the output of energy. The precipitation model achieved regression fitting results of the highest accuracy with training R-value = 0.99976), validation (R-value = 0.99991), and testing (R-value = 0.99931). In addendum, variation in future precipitation and how hydropower plants will generate power in order to adapt to climate change were also addressed in this research. With a mean square error (MSE) of just 1.7262, our suggested model can forecast the production of hydro-electricity generation. The machine learning models created for hydroelectricity generation will minimize operating costs and maximize the energy output of hydropower generation. The suggested remedy is predicated on hydropower plant energy generation forecasts for the future, which can help decision-makers make more informed choices. Given the local weather, the suggested research project will aid in reducing the widening gap between energy output and demand.

Original languageEnglish
Article numbere01981
JournalScientific African
Publication statusPublished - Nov 2023


  • Correlation coefficient
  • Hydo-power generation
  • Machine leaning model
  • Mean Squared Error (MSE)
  • Meteorological variables
  • Regression fitting

ASJC Scopus subject areas

  • Multidisciplinary


Dive into the research topics of 'Hydro-power generation forecast in South Africa based on Machine Learning (ML) models'. Together they form a unique fingerprint.

Cite this