Unveiling Tomorrow's Energy Demand - Enhancing Electricity Consumption Forecasts in Somalia with Linear Regression and Multilayer Perceptron (MLP) Neural Networks

Abstract
Numerous benefits can be derived from forecasting electrical energy consumption,
encompassing both economic and environmental aspects. Efficient resource utilization to
meet the current demand, reduced production costs, and minimized CO2 emissions are among
the advantages. Additionally, accurate prediction facilitates the development of effective
power supply strategies, enables financial planning, supports marketing research, and paves
the way for the integration of renewable or alternative energy sources in the near future. Over
the period from 2000 to 2021, Somalia witnessed a remarkable annual growth rate of 61% in
electricity consumption. This trend is expected to persistently drive the demand for power. In
this study, various techniques such as Linear Regression (LR), Gaussian Processes Regression
(GPR), Multi-Layer Perceptron (MLP), and Sequential Minimal Optimization Regression
(SMOreg) were employed to predict future electricity consumption in Somalia. Inputs for the
estimation model included Somalia's Gross Domestic Product (GDP), population, and
historical electrical consumption data spanning from 2000 to 2023. By utilizing the regression
algorithms provided by the open-source WEKA program, a forecast was generated for
Somalia's projected electrical energy consumption from 2024 to 2030. The findings reveal an
upward trend in the amount of electricity consumed at present, as demonstrated by all four
models.
Keywords: Electricity consumption; artificial neural networks; Somalia; Multiple
regression analysis.