| Abstract: |
The increasing penetration of distributed renewable energy sources and non-linear loads into modern power networks has intensified the demand for power electronic converters that can regulate voltage, frequency, and power quality under highly variable operating conditions. Conventional linear controllers such as Proportional-Integral (PI) and Proportional-Integral-Derivative (PID) schemes struggle to maintain acceptable performance when converters face rapid load transients, parameter drift, and stochastic renewable generation profiles typical of smart grids. This empirical paper investigates the application of Artificial Intelligence (AI) techniques, specifically Artificial Neural Networks (ANN) and Reinforcement Learning (RL) agents combined with Particle Swarm Optimization (PSO), for the real-time control and optimization of DC-DC and DC-AC power electronic converters in smart grid applications. A hardware-in-the-loop simulation tested was developed to collect empirical performance data across five operating dimensions: training convergence, total harmonic distortion (THD), transient voltage response, conversion efficiency, and dynamic load-following accuracy. Data collected from 500 simulation trials were statistically analysed and benchmarked against conventional control. Results show that the AI-based controller reduced average THD by 58.6%, improved conversion efficiency by up to 2.6 percentage points, and reduced settling time after step disturbances by approximately 61% relative to the conventional baseline. These findings empirically confirm that AI-based control and optimization can substantially enhance the reliability, efficiency, and power quality of converters operating within smart grid infrastructures, and the relationships identified between controller architecture and performance metrics are further explored in the discussion and reinforced in the concluding synthesis of this paper. |
