Research map / Graph learning / Graph self-supervised learning
Power Grid Anomaly Detection via Hybrid LSTM-GIN Model (Student Abstract)
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a hybrid LSTM-GIN encoder-decoder anomaly detector for power grid cyberattacks, using GIN node signatures over a 123-node grid graph to reach 99.92% accuracy.
Problem
Detecting cyberattack-induced anomalies in power grid sensor readings (output, frequency, current, voltage) in a way that captures both structural and temporal dependencies without overfitting.
Approach
Combine an LSTM encoder, GIN-derived structural node signatures concatenated to the encoded features, and an LSTM decoder that reconstructs next-step values; flag anomalies when MSE exceeds the 0.99 percentile of training reconstruction error. Models are trained per node category to limit overfitting.
Key results
- Hybrid LSTM-GIN model achieves 99.92% accuracy on power grid anomaly detection.
- Model without GIN structural encoding achieves 97.30%.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add GIN structural encoding (ablation) | LSTM anomaly detection model · Power grid anomaly detection | accuracy: 97.30% -> 99.92% | Helped |
Limitations
- Evaluated only on a single simulated 123-node grid with a limited set of attacks.
- Table 1 column headers are ambiguous (F1 and accuracy both appear to report the same values).
Experimental setup
Datasets: Simulated 123-node power grid baseline dataset Simulated power grid anomaly dataset with attacks at 13, 23, 26 seconds
Benchmarks: Simulated power grid anomaly dataset
Metrics: accuracy F1 score
Baselines: Model without GIN
Builds on: Graph Isomorphism Network Long Short-Term Memory
Closest papers by meaning
- PowerGraph: A power grid benchmark dataset for graph neural networks (NeurIPS 2024 · similarity 0.87)
- PF∆: A Benchmark Dataset for Power Flow under Load, Generation, and Topology Variations (NeurIPS 2025 · similarity 0.82)
- Disentangled Table-Graph Representation for Interpretable Transmission Line Fault Location (AAAI 2025 · similarity 0.82)
- Graph Mixture of Experts and Memory-augmented Routers for Multivariate Time Series Anomaly Detection (AAAI 2025 · similarity 0.82)
- PowerPM: Foundation Model for Power Systems (NeurIPS 2024 · similarity 0.81)
- Dissect Black Box: Interpreting for Rule-Based Explanations in Unsupervised Anomaly Detection (NeurIPS 2024 · similarity 0.81)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.