Asian Basic and Applied Research Journal

Electrical faults in distribution networks supplying critical facilities can interrupt operations and damage equipment. This study evaluated a simulation-based artificial intelligence (AI) framework for fault detection and energy management in a renewable-integrated smart grid at the Centre for Skills Development and Training (CSDT). The investigation was informed by the site's 33 kV incoming supply, transformer and Ring Main Unit (RMU) configuration, single-line diagrams, and documented fault events involving a 5 MVA transformer. MATLAB and Python simulations examined normal operation, fault initiation, protective isolation, emergency restoration, ring-circuit restoration and full restoration. Fault-detection performance was assessed using voltage, current, power, frequency and wavelet-derived signal features. Random Forest, support vector machine, artificial neural network and discrete wavelet transform–deep learning models were compared. In the computational evaluation, the proposed model attained 99.1% classification accuracy and an average detection time of 8.7 ms. Under the ring-circuit restoration scenario, the network supplied 3.20 MW of a modelled 3.40 MW demand, corresponding to approximately 94.1% of demand and leaving 0.20 MW unserved. Simulated voltage profiles improved following reconfiguration relative to fault conditions. These findings suggest that combining intelligent fault identification with protective isolation and alternative supply routing may improve supply continuity at the site. As the performance measures originate from simulated scenarios, practical application requires validation of protection coordination, equipment limits and real-time operation.