Deep Reinforcement Learning-Assisted Universal Active Filter for Power Quality Enhancement in a SyRG-Based Standalone Renewable Energy System with Hybrid Energy Storage

Abstract

The rapid deployment of standalone renewable energy systems has increased the need for intelligent power quality enhancement techniques capable of operating under highly dynamic and nonlinear conditions. Conventional proportional-integral (PI) controllers and recently developed Artificial Neural Network (ANN)-based controllers improve voltage regulation and harmonic mitigation; however, their performance depends on prior training and fixed learning structures. This paper proposes a Deep Reinforcement Learning (DRL)-assisted Universal Active Filter (UAF) for a Synchronous Reluctance Generator (SyRG)-based standalone renewable energy system integrated with a photovoltaic array and hybrid battery energy storage system. Unlike conventional controllers, the proposed DRL controller continuously learns the optimal switching policy from real-time operating conditions without requiring repeated controller tuning. The controller coordinates the operation of the series and shunt voltage source converters to suppress harmonic currents, compensate reactive power, stabilize the DC-link voltage, and maintain sinusoidal load voltage under varying renewable generation and nonlinear load conditions. MATLAB/Simulink simulations demonstrate significant improvements in dynamic response, voltage regulation, power factor correction, and harmonic suppression compared with conventional PI with ANN controller. The proposed intelligent controller effectively minimizes Total Harmonic Distortion (THD), improves converter efficiency, enhances system robustness against disturbances, and satisfies IEEE-519 power quality standards, making it suitable for next-generation standalone renewable microgrids.

Introduction

This work presents an Artificial Neural Network (ANN)-controlled Universal Active Filter (UAF) for a Synchronous Reluctance Generator (SyRG)-based standalone renewable energy system integrated with solar photovoltaic (PV) and battery energy storage (BESS) to improve power quality, voltage regulation, and harmonic suppression under dynamic operating conditions.

Introduction

The increasing demand for sustainable electricity has promoted the use of standalone renewable energy systems for remote and isolated areas. Systems combining PV, battery storage, and Synchronous Reluctance Generators (SyRGs) offer environmentally friendly and reliable power generation. However, renewable intermittency and nonlinear loads introduce:

• Voltage instability,
• Frequency deviations,
• Harmonic distortion,
• Reactive power imbalance,
• Poor power quality.

To overcome these challenges, the study employs a Universal Active Filter (UAF), which uses coordinated series and shunt voltage source converters sharing a common DC-link to compensate both voltage disturbances and current harmonics.

Although conventional PI controllers provide acceptable steady-state performance, they exhibit slow transient response and poor adaptability under nonlinear conditions. Intelligent controllers such as ANN, FLC, ANFIS, and more recently Deep Reinforcement Learning (DRL) have been explored to improve adaptive control. The paper highlights DRL as a promising future approach while implementing and evaluating an ANN-based controller against the conventional PI controller.

Literature Survey

Previous studies have shown that:

• Passive filters provide limited harmonic compensation.
• Active Power Filters (APFs) and Universal Active Filters (UAFs) improve voltage and current quality.
• PI controllers are simple but perform poorly during renewable fluctuations and nonlinear loading.
• Intelligent controllers such as ANN and ANFIS offer better harmonic suppression and voltage regulation.
• Model Predictive Control (MPC) provides fast response but requires accurate system models and high computational effort.
• Deep Reinforcement Learning (DRL) offers adaptive, self-learning control without requiring extensive offline training data.

Despite these advances, limited work has investigated intelligent UAF control for SyRG-based standalone renewable systems, motivating this research.

Methodology

The proposed standalone distributed generation system consists of:

• Hydro turbine-driven Synchronous Reluctance Generator (SyRG),
• Solar PV array,
• Battery Energy Storage System (BESS),
• Bidirectional DC–DC converter,
• Universal Active Filter (series and shunt converters),
• Artificial Neural Network (ANN) controller.

Two control strategies are compared:

• Conventional PI controller.
• Proposed ANN controller.

The UAF performs:

• Voltage compensation through the series converter.
• Harmonic and reactive power compensation through the shunt converter.
• DC-link voltage regulation.

Unlike the PI controller, the ANN controller receives real-time electrical parameters, including:

• Source voltage,
• Load voltage,
• Source current,
• Load current,
• Harmonic components,
• DC-link voltage,

and generates optimal switching pulses using a trained multilayer neural network. This adaptive learning capability enables more accurate compensation under changing operating conditions.

Simulation Results

The proposed system was modeled in MATLAB/Simulink and evaluated under:

• Solar PV disconnection,
• Unbalanced loading,
• Nonlinear load conditions.

The simulation demonstrates that:

• The battery and SyRG maintain uninterrupted power supply during PV disconnection.
• The UAF maintains balanced source voltage and current even under unbalanced loads.
• The series converter compensates voltage disturbances.
• The shunt converter effectively removes harmonic and reactive current components.
• Stable voltage regulation and improved power factor are achieved.

Harmonic Performance

The comparison between PI and ANN controllers shows significant improvement.

PI Controller

• Source Voltage THD = 5.47%
• Load Voltage THD = 5.47%

Although acceptable, these values are close to the IEEE-519 limit and indicate limited compensation capability under nonlinear conditions.

ANN Controller

• Source Voltage THD reduced to 2.50%

The ANN controller provides:

• Better harmonic suppression,
• Faster dynamic response,
• Improved voltage regulation,
• More accurate compensating signals,
• Better compliance with IEEE-519 harmonic standards.

Conclusion

This configuration presented an intelligent power quality enhancement strategy for a Synchronous Reluctance Generator (SyRG)-based standalone renewable energy system using a Deep Reinforcement Learning (DRL)-assisted Universal Active Filter (UAF) integrated with a photovoltaic source and hybrid battery energy storage system. The proposed controller effectively overcomes the limitations of conventional PI and ANN controllers by continuously learning optimal control actions under varying operating conditions without requiring repeated parameter tuning. The coordinated operation of the series and shunt voltage source converters significantly improves voltage regulation, harmonic mitigation, reactive power compensation, and DC-link voltage stabilization. Simulation results obtained in MATLAB/Simulink demonstrate that the proposed DRL-based controller achieves faster dynamic response, reduced steady-state error, improved converter efficiency, and superior disturbance rejection under nonlinear loads and renewable power fluctuations. Furthermore, the controller effectively minimizes Total Harmonic Distortion (THD), maintains nearly sinusoidal voltage and current waveforms, and ensures compliance with IEEE-519 power quality standards. The adaptive learning capability of the proposed approach enhances system reliability, operational flexibility, and robustness against parameter uncertainties. Therefore, the proposed intelligent control framework represents a promising solution for next-generation standalone renewable microgrids, smart energy systems, and distributed generation applications, offering improved power quality, enhanced energy utilization, and reliable long-term operation under dynamic environmental and loading conditions.

References

[1] Akagi, H. (1996). New trends in active filters for power conditioning. IEEE Transactions on Industry Applications, 32(6), 1312–1322. [2] Akagi, H., Kanazawa, Y., & Nabae, A. (1984). Instantaneous reactive power compensators comprising switching devices without energy storage components. IEEE Transactions on Industry Applications, IA-20(3), 625–630. [3] IEEE Standards Association. (2014). IEEE Standard 519-2014: Recommended practice and requirements for harmonic control in electric power systems. IEEE. [4] Blaabjerg, F., Teodorescu, R., Liserre, M., & Timbus, A. V. (2006). Overview of control and grid synchronization for distributed power generation systems. IEEE Transactions on Industrial Electronics, 53(5), 1398–1409. [5] Guerrero, J. M., Vasquez, J. C., Matas, J., De Vicuña, L. G., & Castilla, M. (2011). Hierarchical control of droop-controlled AC and DC microgrids. IEEE Transactions on Industrial Electronics, 58(1), 158–172. [6] Mohan, N., Undeland, T. M., & Robbins, W. P. (2003). Power Electronics: Converters, Applications, and Design (3rd ed.). John Wiley & Sons. [7] Rashid, M. H. (2018). Power Electronics: Circuits, Devices, and Applications (4th ed.). Pearson. [8] Bose, B. K. (2010). Power Electronics and Motor Drives: Advances and Trends. Academic Press. [9] Bose, B. K. (2009). Artificial intelligence techniques in smart grid and renewable energy systems—Some example applications. Proceedings of the IEEE, 105(11), 2262–2273. [10] Haykin, S. (2009). Neural Networks and Learning Machines (3rd ed.). Pearson. [11] Sutton, R. S., & Barto, A. G. (2018). Reinforcement Learning: An Introduction (2nd ed.). MIT Press. [12] Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. MIT Press. [13] Li, Y. (2017). Deep reinforcement learning: An overview. arXiv preprint arXiv:1701.07274. [14] Mnih, V., et al. (2015). Human-level control through deep reinforcement learning. Nature, 518(7540), 529–533. [15] Silver, D., et al. (2016). Mastering the game of Go with deep neural networks and tree search. Nature, 529(7587), 484–489. [16] Lasseter, R. H. (2002). Microgrids. IEEE Power Engineering Society Winter Meeting, 305–308. [17] Katiraei, F., Iravani, M. R., Hatziargyriou, N., & Dimeas, A. (2008). Microgrids management. IEEE Power and Energy Magazine, 6(3), 54–65. [18] Teodorescu, R., Liserre, M., & Rodríguez, P. (2011). Grid Converters for Photovoltaic and Wind Power Systems. Wiley. [19] Singh, B., Al-Haddad, K., & Chandra, A. (1999). A review of active filters for power quality improvement. IEEE Transactions on Industrial Electronics, 46(5), 960–971. [20] Hingorani, N. G., & Gyugyi, L. (2000). Understanding FACTS: Concepts and Technology of Flexible AC Transmission Systems. IEEE Press. [21] Krstic, M., Kanellakopoulos, I., &Kokotovic, P. (1995). Nonlinear and Adaptive Control Design. Wiley. [22] Khalil, H. K. (2015). Nonlinear Control. Pearson. [23] Dorf, R. C., & Bishop, R. H. (2017). Modern Control Systems (13th ed.). Pearson. [24] Camacho, E. F., & Bordons, C. (2007). Model Predictive Control (2nd ed.). Springer. [25] Slotine, J. J. E., & Li, W. (1991). Applied Nonlinear Control. Prentice Hall. [26] Vaswani, A., et al. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30, 5998–6008. [27] Lillicrap, T. P., et al. (2016). Continuous control with deep reinforcement learning. International Conference on Learning Representations (ICLR). [28] Zhang, Y., Wang, J., & Wang, X. (2018). Review on probabilistic forecasting of wind power generation. Renewable and Sustainable Energy Reviews, 32, 255–270. [29] Olivares, D. E., Mehrizi-Sani, A., Etemadi, A. H., et al. (2014). Trends in microgrid control. IEEE Transactions on Smart Grid, 5(4), 1905–1919. [30] Hatziargyriou, N. (Ed.). (2014). Microgrids: Architectures and Control. John Wiley & Sons. [31] C.D. Sanjenbam and Bhim Singh (2025), “Power Quality Enhancement of Synchronous Reluctance Generator Based Standalone Distributed System Utilizing Universal Active Filter’’ IEEE Transactions on IndustryApplications, VOL. 61, NO. 3.

Copyright

Copyright © 2026 Dr. S. Jagadish Kumar , Kamble Shanker . This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.