World Journal of Engineering Research and Technology

*ThankGod Sylvanus Ntem, Eko James Akpama, Peter Ohiero O., Iwueze Ifeanyi Moses.

ABSTRACT

This study presents a novel approach to mitigating peak-demand-induced disruptions in power distribution networks through the integration of solar distributed generation (DG) with an artificial neural network (ANN) predictive controller. Addressing a critical gap in existing grid response strategies which remain centralized and reactive, the study develops a data-driven control architecture combining ETAP-based load flow analysis with MATLAB/Simulink simulations of a 217.12 kW solar DG system. A Bayesian regularized ANN trained on historical load profiles (residential, commercial, industrial) achieves 92% accuracy in predicting peak windows (09:00–18:00), enabling proactive DG dispatch. During peak demand, the DG supplies 200 kW (33% of total load), reducing grid dependence from 600 kW to 400 kW and cutting service disruptions by 50%. Voltage fluctuations improve from ±10% to ±5%. Unlike conventional demand response, this solution maintains grid stability without load rescheduling, which iscritical for industrial users. Validated on Nigeria’s Cross River State Waterboard (CRSWB) distribution network, this work demonstrates how machine learning-enhanced DG integration can transform passive distribution systems into resilient smart grids. The methodology is scalable to other regions with unreliable centralized generation, offering a blueprint for energy transition in developing economies.

[Full Text Article] [Download Certificate] https://doi.org/10.5281/zenodo.17490280