A genetic programming-based ensemble method for long-term electricity demand forecasting.

Area of Science:

• Energy Economics
• Computational Intelligence
• Forecasting Science

Background:

• Accurate long-term electricity consumption forecasting is crucial for Ethiopia's energy infrastructure planning.
• Existing forecasting models may not capture the complex dynamics of Ethiopia's growing energy demand.

Purpose of the Study:

• To develop and validate a novel ensemble method for long-term electricity consumption forecasting in Ethiopia.
• To project electricity consumption trends up to the year 2031.

Main Methods:

• A two-stage ensemble approach was employed, integrating genetic algorithms (GA), particle swarm optimization (PSO), and simulated annealing (SA) with regression models (linear, quadratic, exponential).
• Genetic programming (GP) was utilized in the second stage to refine preliminary forecasts, creating a final predictive model.
• Performance was evaluated using Mean Absolute Percentage Error (MAPE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R-squared (R²).

Main Results:

• The initial stage yielded accurate forecasts, with GA_Quadratic, PSO_Quadratic, and SA_Quadratic achieving MAPE values of 3.61%, 3.63%, and 4.68%, respectively.
• The final genetic programming-based ensemble model achieved a superior MAPE of 2.83%.
• The proposed model outperformed all first-stage methods across all evaluated error metrics.

Conclusions:

• The novel genetic programming-based ensemble method provides a highly accurate approach for forecasting long-term electricity consumption.
• Ethiopia's electricity consumption is projected to triple by 2031 compared to 2021 levels under various scenarios.
• This forecasting model can significantly aid in strategic energy planning and infrastructure development for Ethiopia.