Purpose This study aims to assess how advanced machine learning can enhance renewable energy forecasting accuracy in Kenya, reduce dependence on expensive diesel backup, lower electricity costs and cut carbon emissions. It compares modern machine learning models against conventional methods, quantifies economic and environmental benefits, identifies major adoption barriers in the Kenyan context and proposes practical, immediately actionable strategies for successful integration into national grid operations. Design/methodology/approach A Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)-guided systematic literature review was conducted. Fifty-three peer-reviewed empirical studies published between 2020 and 2025, with core focus on 2023 to 2025, were selected from Science Direct, Springer Link, Emerald Insight, Web of Science and Wiley Online Library using targeted Boolean searches centred on machine learning and renewable energy forecasting in Kenya and similar tropical environments. Findings Hybrid deep learning models (extreme gradient boosting (XGBoost), long short-term memory (LSTM), convolutional (CNN) and transformer architectures) consistently outperform traditional persistence, ARIMA and global numerical weather prediction models. They reduce forecast errors by 35 to 52%, deliver large diesel fuel savings and substantially cut CO2 emissions. Main barriers are poor grid telemetry, high upfront costs and scarce local expertise. Immediate solutions include open data platforms, short intensive training and focused pilots at major renewable sites. Research limitations/implications The study relies on systematic review of existing literature rather than new primary Kenyan data collection. Rapid evolution of machine learning techniques means some models may soon be surpassed. Findings remain broadly applicable across Sub-Saharan Africa but require local validation and continuous updating as more Kenyan telemetry becomes available. Practical implications Kenya Power, KenGen and independent power producers should establish national real-time renewable data-sharing platforms, launch intensive engineer training bootcamps and run forecasting pilots at Olkaria, Lake Turkana and Garissa. Regulators can accept probabilistic machine learning forecasts for reserve calculations and offer incentives for Forecasting-as-a-Service contracts, quickly reducing diesel use and consumer tariffs. Social implications Accurate forecasting will minimise outages, stabilise electricity supply in rural areas, lower household and business energy costs, reduce air pollution from diesel generators and help Kenya meet Sustainable Development Goal 7 on affordable and clean energy while advancing energy equity and environmental justice for all citizens. Originality/value This is the first PRISMA-guided systematic review focused exclusively on applying state-of-the-art machine learning forecasting improvements to Kenya’s unique renewable-heavy grid. It provides policymakers and utilities with an evidence-based, immediately actionable roadmap to convert Kenya’s abundant renewable resources into reliable, affordable and truly clean electricity.
Morara et al. (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: