Data analytics for prediction of solar PV power generation and system performance: A real case of Bui Solar Generating Station, Ghana

The grid-connected solar power generated by the Bui Power Authority is sold to Ghana Grid Company Limited (GRIDCo) and other customers through bilateral contracts. However, there have been challenges in meeting the supply commitments due to the stochastic nature of solar energy. Fluctuating weather and climatic conditions make it difficult for operators to predict the output of the solar photovoltaic (PV) plant in advance. This has the tendency of leading to anticipated power loss or excess power not being managed effectively. In this study, the random forest and gradient boosting regressor algorithms were used to produce deterministic and prob-abilistic predictions of solar power generation by using data collected over an eleven-month span. Some of the findings show that the random forest model that produced probabilistic predictions performed better than the other compared models in terms of its accuracy. It produces reliable predictions with a normalized mean absolute error of 1.18%. This is deemed acceptable for operational purposes in the sphere of renewable energy prediction. As a result, these predictions are efficient enough to be used by utility companies as input to numerous decision-making problems during operations which consequently leads to the grid being managed effectively. It may also help them to choose the appropriate modelling approach for predicting solar PV power. Overall, this study supports the generation of clean and reliable energy through improved solar power prediction which directly contributes to African Union's agenda 2063 goals especially, in creating environmentally sustainable and climate resilient economies and communities. This is directly linked to the 7th and 13th UN sustainable developmental goals in ensuring access to affordable, reliable, sustainable, and modern energy for all and, combating climate change and its impacts.