Safe Deep Reinforcement Learning for Resilient Self-Proactive Distribution Grids Against Wildfires
Bibliographic record
Abstract
With the growing risks and frequency of wildfires, power distribution systems (PDS) face significant challenges in maintaining reliability and security. Existing literature primarily focuses on post-event service restoration using stochastic optimization methods. Nevertheless, such approaches fall short in managing the dynamic and uncertain nature of wildfires, particularly when it comes to taking proactive measures to mitigate power outages. To address this problem, this paper introduces a wildfire smart resilience controller (WF-SRC) that utilizes a model-assisted safe Deep Reinforcement Learning (DRL) mechanism to reduce the impacts of wildfire-induced disruptions. The WF-SRC continuously monitors and analyzes both the status of the PDS and the spatiotemporal dynamics of wildfires, then executes preemptive actions to prevent wildfires from compromising distribution lines. These actions include optimally dispatching stationary and mobile distributed energy resources (DERs) that operate under a master-slave control scheme. While recent works assume full observability and formulate the PDS resilience problem as a Markov Decision Process (MDP), this approach leads to an impractically large state space and limited realism. In contrast, our approach models the problem as a Partially Observable Markov Decision Process (POMDP). This explicitly captures real-world sensing limitations, such as noisy measurements that arise during extreme events. The POMDP is tackled using an LSTM-TD3 DRL agent, enabling effective sequential decision-making in environments with incomplete information. Using real-world data from Alberta wildfires, simulation results demonstrate the effectiveness of the proposed solution in reducing wildfire impact, optimizing energy distribution, and enhancing robustness to uncertainties, ultimately contributing to a more resilient and proactive power grid.
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 | 0.002 |
| 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 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 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".