Jul 2026· 2026 4th International Conference on Sustainable Computing and Smart Systems (ICSCSS)· pp. 488-493· 0 citations· 21 references
Abstract
The integration of distributed energy resources into active distribution networks presents fundamental challenges for voltage regulation, with voltage violations affecting 23% of distribution feeders under high renewable penetration scenarios. Existing approaches, notably MADDPG and PPO-based methods, address multi-agent coordination but fail to account for data privacy constraints in distributed control architectures, resulting in a 31% performance degradation under communication constraints. This work addresses this gap by introducing FedDRL-VC, a federated deep reinforcement learning framework for decentralized voltage control. FedDRL-VC employs a hierarchical aggregation mechanism to preserve data locality while enabling collaborative policy learning across network zones. A priority experience replay mechanism was designed to accelerate convergence on critical voltage events. Training was conducted on the IEEE 123-bus test feeder with realistic DER profiles over 10,000 episodes. FedDRL-VC achieved a voltage deviation of 2.1% on the benchmark, surpassing MADDPG by 56% (p < 0.001, Cohen's d = 1.84). Computational cost was reduced by 65%; convergence in 320 iterations versus 890 for TD3. Privacy preservation was maintained with zero data exchange between agents.
With the increasing penetration of distributed generation in active distribution networks, voltage fluctuation and voltage violation problems have become more prominent, especially in weak grids. Conventional voltage control methods based on empirical parameter tuning or single-device regulation often have limited adaptability under time-varying load conditions and fluctuating distributed generation output. To address these issues, this paper proposes a deep reinforcement learning based coordinated voltage control method for HST and DG. The proposed method takes monitored node voltages, load levels, and distributed generation outputs as state inputs and uses the gain adjustments of HST and DG as control actions, thereby achieving adaptive optimization of voltage response through a continuous decision-making framework. On this basis, an experimental validation framework including training and validation datasets, multi-strategy comparison communication-constrained analysis, and repeated-run statistics is established to evaluate the control performance generalization capability and robustness of the proposed method. The results show that the proposed strategy outperforms the baseline method in voltage deviation, overshoot oscillation energy, and control effort, while maintaining good performance consistency in the validation scenario. Under mild communication constraints, the proposed strategy still exhibits acceptable adaptability and operational stability. These findings indicate that deep reinforcement learning provides an effective optimization approach for coordinated HST and DG control and offers a useful reference for voltage regulation in active distribution networks.
High levels of photovoltaic (PV) generation in distribution networks create substantial uncertainty and voltage variability, which limits the effectiveness of conventional deterministic distribution network reconfiguration (DNR) strategies. In PV-dominated feeders, rare but severe operating conditions may considerably influence active power losses and voltage stability. To address this challenge, this paper proposes a risk-informed optimization framework for DNR that combines reinforcement learning with probabilistic performance assessment. A Deep Q-Network (DQN) agent is designed to support the selection of feasible radial switching configurations by interacting with the distribution network environment. Throughout the learning process, candidate network topologies are evaluated through radial load flow calculations, while a composite objective function incorporating active power losses and voltage deviation steers the agent toward improved configurations. The training stage is based on deterministic performance indices; however, the final reconfiguration solution is assessed under uncertainty to examine its operational robustness. For this purpose, extensive Monte Carlo simulations are performed to capture the stochastic behavior of PV generation and load demand. Tail-based risk metrics, including Value at Risk (VaR) and Conditional Value at Risk (CVaR), are computed for both loss and voltage deviation indices, providing insight into the performance of the selected configuration under unfavorable operating scenarios. The proposed framework is first validated on the IEEE 33-bus distribution system and then further investigated on the IEEE 69-bus network. The obtained results demonstrate that the proposed DQN-based reconfiguration approach can enhance voltage profiles and reduce power losses under high PV penetration. In addition, the probabilistic analysis identifies meaningful trade-offs between efficiency and voltage robustness, highlighting the importance of considering uncertainty-driven risk assessment in computational decision-making for modern active distribution networks.
A hierarchical autonomous control framework featuring large language model-driven dynamic reward shaping (LLM-Driven DRS) is introduced to balance security and efficiency and achieves Pareto superiority over conventional static weight strategies.
The digital transformation of modern power grids demands intelligent, data-driven strategies for long-term investment decision-making under uncertainty. Traditional deterministic or rule-based optimization approaches struggle to adapt to stochastic market dynamics, renewable intermittency, and evolving operational constraints. This paper presents a Deep Reinforcement Learning–Enhanced Dynamic Optimization Framework (DRL-DOF) that integrates uncertainty-aware policy optimization, temporal-attention actor–critic networks, and digital twin simulation for precision investment management. The proposed method formulates grid investment as a constrained Markov Decision Process, balancing return, cost, and risk via a Conditional Value-at-Risk (CVaR)-based reward function. A federated learning mechanism further enables decentralized coordination across regional grids without compromising data privacy. Experimental results across synthetic and real datasets demonstrate that DRL-DOF achieves up to 15% higher cost efficiency, enhanced reliability, and faster convergence than state-of-the-art optimization and baseline DRL methods. This work establishes a scalable and interpretable foundation for intelligent investment decision-making in sustainable and resilient power systems.
Yu-Hao Zhou· International Conference on...· 0 citations
Microgrids play a critical role in enhancing the flexibility, reliability, and sustainability of modern power systems by integrating distributed energy resources, energy storage systems, and controllable loads. However, the inherent uncertainty of renewable generation and the stochastic nature of load demand pose significant challenges to optimal energy management. To address these issues, this paper proposes a deep reinforcement learning (DRL)-based optimal energy management framework for microgrids. The problem is formulated as a Markov decision process, where the system state captures renewable generation, load demand, and storage status, while the control actions determine power dispatch and energy storage operation. A deep reinforcement learning model is developed to learn optimal control policies through continuous interaction with the environment, enabling adaptive decision-making under dynamic and uncertain conditions. To improve learning efficiency and policy stability, state normalization and reward shaping strategies are incorporated. Furthermore, a constrained optimization mechanism is introduced to ensure operational safety and economic feasibility. Experimental results on benchmark microgrid scenarios demonstrate that the proposed method outperforms conventional rule-based strategies and model-based optimization approaches in terms of operational cost reduction, energy utilization efficiency, and robustness under uncertainty. The results indicate that the proposed DRL-based framework provides an effective and scalable solution for intelligent microgrid energy management.
Li Chen, Hong-Qiao Li, Zhen-Xing Chen et al.· European Conference on Elect...· 0 citations
Convergence and multi-run statistical analysis further confirm the robustness, stability, and reproducibility of the trained policy, demonstrating the effectiveness of DRL as an intelligent and scalable solution for next-generation microgrid PQ control.
Pratibha V. Hurkadli, G. A. Kumar, T. Manjunath· Advances in Data Science and...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
