Physics-Guided Multi-Agent Deep Reinforcement Learning for Robust Active Voltage Control in Electrical Distribution Systems
Bibliographic record
Abstract
Although several multi-agent deep reinforcement learning (MADRL) algorithms have been employed in power distribution networks configured with high penetration level of Photovoltaic (PV) generators for active voltage control (AVC), the impact of the voltage fluctuation of a single PV node on voltage violations of other PV nodes in the network is ignored. Consequently, it leads to the conservativeness of the existing MADRL based AVC algorithms. In this paper, a robust MADRL control algorithm is designed to minimize the nodal voltage violation and line loss with the exploration of coupling voltage fluctuations across all the controlled nodes by coordinating PV inverters, and a physics factor is utilized to guide (physics-guided) the training policy with the expectation of a better performance compared to existing purely data-driven methods. In the proposed physics-guided multi-agent adversarial twin delayed deep deterministic (PG-MA2TD3) policy gradient algorithm, a physics factor, global sensitivity of voltage (GSV), is properly embedded in the algorithm to measure the influence of the nodal voltage fluctuation on voltage violations on the other controlled nodes with PV inverters and this GSV is shared in the learning center to guide the centralized learning and decentralized execution process. The multi-agent adversarial learning (MAAL) embedded with the GSV to seek an adaptive descend gradient for reducing the Q-value function appropriately rather than always assuming the worst case. Therefore, this physics-guided method can reduce the conservation and provide significantly better reward. Finally, the proposed algorithm is compared with several other methods on IEEE 33-bus, 141-bus and 322-bus with three-year data in Portuguese and the results indicate the proposed method can obtain the minimal voltage fluctuation and the best reward in the comparisons.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".