Electric power is a unique product that is generated, delivered and consumed at the same time requires dynamic monitoring. In smart era, widespread use of sensitive and nonlinear loads driven by power electronics, low power factor and system events like faults are susceptible to poor power quality and power interruptions. Power quality (PQ) problems refer to issues with voltage, current, or frequency that lead to interruptions or equipment failures. PQ problems such as voltage sag, unbalance, and harmonic distortion have been recorded at Bahir Dar Textile Share Company (BTSC), leading to economic losses from equipment damage and working time. A series-connected, Dynamic Voltage Restorer (DVR) system is a custom power device that provides a solution with cost-effectiveness, compact size, and rapid response to voltage disturbances, effectively mitigating power quality issues through optimized and adaptive techniques. SVPWM-based VSI converter for the DVR is chosen for its faster response and reduced switching losses, followed by an LCL filter optimizing THD with the Firefly algorithm for a smooth sinusoidal output. The input-output data for an artificial neural network (ANN) was trained after tuning PID gain parameters using particle swarm optimization (PSO) technique. By intentionally creating faults such as single-line to ground (SLG), double-line to ground (DLG), threephase to ground (TPG), and multi-stage (MSF) faults, along with associated harmonics, a significant voltage sag occurred, causing reductions in rms voltage in phase A, phases A and C, and phases A, B, and C voltages, respectively. This thesis aims to mitigate power quality at the BTSC New Spinning section by using PSO-tuned PID gain parameters for an ANN-controlled DVR and addressing voltage sags, harmonics, and imbalances to improve the quality of sensitive loads through simulations in a MATLAB/Simulink environment. The ANN-based DVR significantly enhanced per unit rms voltage in SLG, DLG, TLG, and MSF to 0.983, 0.981, 0.985, and 0.987, compared to a system with a PID-based DVR of 0.81, 0.801, 0.817, and 0.805, respectively. Additionally, ANN-based DVR lowered the THD values during SLG, DLG, TLG, and MSF 2.58%, 1.69%, 1.54%, and 2.4%, respectively, in contrast to a system with a PID-based DVR 5.76%, 7.15%, 6.35%, and 7.06%, respectively. Therefore, based on the comparative simulation results, the ANN-based DVR performs more effectively than both the auto-tuned PID-based DVR and the base case for voltage-related issues and harmonics while meeting IEEE 519-1995 standards. Keywords: BTSC, Dynamic Voltage Restorer, Firefly Algorithm, Power Quality, PSO, SVPWM