Convolutional Variational Autoencoder for Anomaly Detection in On-Load Tap Changers
Bibliographic record
Abstract
Transformer outages significantly impact the reliability and cost efficiency of power systems. Studies indicate that approximately 30% of transformer failures stem from issues with on-load tap changers (OLTC), crucial components in transformer operation. Therefore, continuous monitoring of OLTCs is essential to enhance transformer serviceability. In this study, a vibro-acoustic signal analysis-based monitoring system is employed to assess the condition of OLTCs. This system has been operational since 2016 on three single-phase autotransformers within the Hydro-Québec network, continuously measuring vibration signals from their OLTCs. Notably, these transformers are equipped with sister OLTC units, and the system also records temperature and other pertinent parameters. To detect anomalies in OLTCs and analyze the generated vibration signals, a convolutional variational autoencoder (CVAE) is utilized, trained individually for each transformer family. This approach allows mapping the signal envelope into a two-dimensional latent space using the encoder component of the CVAE, facilitating visual investigation and analysis. The decoder component reconstructs the original input from data in the latent space. Several thresholds based on reconstruction errors are evaluated to detect anomalies, achieving optimal thresholds for each family. This results in anomaly detection rates of 4%, 5%, and 2%, respectively, when tested on data from within the same family not used in the training phase. Furthermore, when tested on data from the other two families, the anomaly detection rates are 99%, 99%, and 100%, respectively. These findings underscore the methodology’s accuracy and effectiveness in identifying anomalies in OLTC operations and distinguishing between different transformer families. Consequently, it holds promise for preemptively identifying potential future anomalies.
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 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 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".