International Journal Evolving Sustainable and Renewable Energy Solutions
AI-Driven Models for Reducing Carbon Emissions in Hybrid Renewable-Non-Renewable Energy Grids: A Nigerian Case Study
Abstract
Idowu Olugbenga Adewumi and Ayoade Oladele Atere
By 2025, Nigeria’s power sector has an installed capacity of 13.4 GW but provides only 5.1 GW, leading to 6.5 hours of load shedding per day, a reliability index of 72%, and yearly CO2 emissions totaling 100 MtCO2. This research formulates and assesses three artificial intelligence (AI) models Random Forest (RF), Long Short-Term Memory (LSTM), and Reinforcement Learning (RL) to enhance grid load allocation and diminish carbon emissions. Analysis of model performance indicated that RF reached RMSE = 210.5 MW, MAE = 145.3 MW, R2 = 0.89, LSTM attained RMSE = 185.2 MW, MAE = 130.7 MW, R2 = 0.92, whereas RL excelled with RMSE = 178.4 MW, MAE = 122.9 MW, R2 = 0.94. In optimized scenarios, RF cut annual CO2 emissions from 100 MtCO2 to 82 MtCO2 (an 18% reduction), LSTM to 78 MtCO2 (a 22% reduction), and RL to 74 MtCO2 (a 26% reduction). The use of renewables rose from a baseline of 28% to 34% (RF), 36% (LSTM), and 39% (RL). Grid reliability enhanced to 79%, 82%, and 85%, while load shedding decreased to 5.0, 4.6, and 4.1 hrs/day, correspondingly. Sensitivity analysis indicated a +10% renewable share resulted in CO2 reductions of 19.6%, 21.7%, and 23.4% across the three models. National forecasts based on RL optimization indicate that by 2030, the share of renewables will reach 45%, emissions will decrease to 72 MtCO2/year (a 28% reduction), and reliability is projected to rise to 86%, resulting in an annual avoidance of about 9.6 MtCO2. Comparative benchmarking shows Nigeria’s RL-optimized results (39% renewables, 26% CO2 reduction) surpassing South Africa (32%, 18%) but trailing behind Kenya (48%, 30%) and Germany (55%, 35%). Adherence to Nigeria’s Energy Transition Plan (ETP-2030) shows 150% achievement on renewable goals, 140% achievement on CO2 emission reduction, 108% achievement on reliability, and 162% achievement on reducing load-shedding. These results validate that AI-based optimization provides quantifiable improvements in accuracy, efficiency, and sustainability, aiding Nigeria’s low-carbon shift.