Power quality improvement in Grid‐Connected PV system using fractional order controller and Fractional-order Lipschitz Recurrent Neural Network (FLRNN)

Abstract

In hybrid renewable energy systems (HRES), particularly with solar, fuel cell, and battery components, common PQ disturbances that occur are voltage sags, swells, and fluctuations. An intelligent FLRNN-FOPID-DSTATCOM control framework is proposed that integrates a Fractional-Order PID (FOPID) controller, whose parameters are optimized using a novel Draft-Mongoose Tailored Earthworm Optimizer (DTEO), and a Fractional-Order Lipschitz Recurrent Neural Network (FLRNN), for PQ improvement under varying load and source conditions. The simulated experimental setup used the MATLAB/Simulink environment, after which this approach underwent a rigorous comparative study with traditional PID, Meta-heuristic PI/PID, and Sliding Mode Controllers (SMC). The quantitative nature of the results proves a substantial reduction of THD down to 0.0043 from undefined baseline THD values, signifying very good harmonic suppression. The system is able to stabilize the PV voltage and current from varying ranges of −200 V to 350 V and −800 A to 300 A, respectively, to steady outputs of 350 V and 300 A. While the battery and DC-link voltages are restored from momentary dips as low as 200 V to steady state voltages of 270–300 V, variability due to voltage sags, swells, and fluctuations are suppressed such that LV is stabilized within ±2.5–3 V and IC ramped effectively to ±500–598 A. Therefore, compared to the existing methods, the proposed controller greatly reduces THD more swiftly with better dynamic response and adaptability to PV variability, saving times required for the neural network training. From the above results, the proposed technique can be considered as a promising high-performance method to enhance power quality in real-time in a grid-connected HRES environment.