Skip to content

Data-Driven Adaptive Reactive Power and Voltage Control Method for Distribution Networks with Energy Storage by Using DRL-SAC Algorithm

Sep 2026 · Processes · 0 citations · 23 references
Microgrid Control and Optimization

Abstract

In distribution network and microgrids, energy storage (ES) systems possess four-quadrant operational capabilities, making them inherently high-quality resources for reactive power (RP) regulation. However, existing research has primarily focused on optimizing the active power of ES to achieve economic objectives, while the potential for RP and voltage control has not been fully explored. Meanwhile, traditional RP optimization methods have inherent limitations in terms of real-time performance, addressing uncertainty, and handling nonlinear problems. To address these challenges, this paper proposes a data-driven adaptive RP and voltage control method for distribution network with ES using deep reinforcement learning (DRL)–Soft Actor–Critic (SAC) algorithm. First of all, this method models the grid’s RP and voltage control problem as a sequential decision-making process, with the core being the construction of a control agent that integrates grid operational states with a deep neural network. Through continuous interaction with the environment, this agent autonomously learns and dynamically adapts to the random fluctuations in photovoltaic (PV) output and load without relying on precise physical models. Secondly, this paper sets minimizing network losses, voltage deviations, and the operational costs of RP equipment in ES as comprehensive optimization objectives, translating them into a reward function within the DRL-SAC framework. Leveraging the powerful nonlinear mapping capabilities and extremely fast forward computation speed of deep neural networks, the strategy achieves a data-driven approximation of the optimal RP control strategy in complex grid environments. Finally, the superiority of the strategy is comprehensively verified on the modified IEEE 33-bus system under three typical operating conditions (daytime fluctuation, extreme weather, sudden load change). The results show that the voltage qualification rate is increased to 99.1% and the network loss is reduced by 33.7%, providing an engineering-feasible solution for ES systems to participate in distribution network RP and voltage regulation.

Read PDF

Similar papers

Open access 2026

Real-Time Energy Management of PV-ESS Integrated Active Distribution Networks Using Digital Twin-Enabled Deep Reinforcement Learning

The proposed DT-DRL framework establishes a closed-loop cyber–physical architecture in which continuously synchronized Digital Twin states are directly incorporated into a Proximal Policy Optimization (PPO)-based decision-making process to jointly minimize operating cost, voltage deviation, and battery degradation whil...

M. Le · 0 citations
Open access Sep 2026

Data-Driven Fault Ride-Through Operation of Distributed Grid-Forming Inverters Using Multi-Agent Reinforcement Learning

A multi-agent reinforcement learning inspired technique for the parameter selection of virtual synchronous generator (VSG)-controlled GFMIs through independent twin delayed deep deterministic policy gradient (TD3PG) agents, considering symmetrical and asymmetrical grid faults on the IEEE 13 bus network is proposed.

S. Chand, Arman Ali, S. A. Ali et al. · 0 citations
Open access Oct 2026

Voltage support strength-constrained transmission network expansion planning for renewable-integrated grids via deep reinforcement learning

A novel planning framework that integrates data-driven uncertainty modeling with intelligent, stability-aware decision-making is introduced that provides a powerful and forward-looking paradigm for TNEP, effectively balancing economic, renewable accommodation, and dynamic stability objectives to foster a more resilient...

Jia-Ming Li, Ke Wang, Yan Li et al. · 0 citations
Aug 2026

Synergistic Dual-Loop Control for Energy Management and Voltage Stability in PV-BESS Micro-grids

The integration of volatile renewable energy sources, such as photovoltaic (PV) systems, alongside battery energy storage systems (BESS) into DC micro-grids presents significant control challenges, primarily in ensuring DC bus voltage stability and optimizing long-term energy management. This paper introduces a novel s...

H. Chabana, I. Tegani, Salem Tegani et al. · 0 citations
Open access Aug 2026

Coordinated multi-layer voltage control with predictive tap scheduling for active distribution networks

Active distribution networks with high penetration of PV, BESS, and EV charging face significant voltage regulation challenges and accelerated OLTC wear. This paper proposes a coordinated multi-layer voltage control framework operating across multiple time scales. It integrates fast local fuzzy Volt–Var and state-of-ch...

Weverson dos Santos Cirino, T. Soares, I. Torné · 0 citations
Sep 2026

Decentralized power allocation with integrated bus voltage recovery for hybrid energy storage systems

Abstract A hybrid energy storage system (HESS) plays a crucial role in stabilizing DC microgrids against power fluctuations from renewable sources and loads. To mitigate severe bus voltage deviations under complex disturbances, this paper proposes a composite control method integrating a super-twisting sliding mode con...

Shu-Fan Wang, Bo Wei · 0 citations

Related blog posts

MIT News · Artificial Intelligence Nov 17, 2026

Learning the language of machine learning

By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.

Microsoft Research Blog Nov 16, 2026

TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient

AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.