[POSTER] HierarchyScope: Spatio-Temporal Data Summarization for Efficient Voltage Forecasting and Anomaly Detection in Hierarchical Power Grids
Abstract
This repository contains the poster "HierarchyScope: Spatio-Temporal Data Summarization for Efficient Voltage Forecasting and Anomaly Detection in Hierarchical Power Grids". The work is presented at the 3rd SESBC Conference 2025 in Uppsala.
Full text
Dynamically adding queries to production monitoring at no cost. The Problem: A Data-Rich, Insight-Poor Grid: •Data Overload:Smart meters (AMI) and PMUs generate massive, high-velocity data streams. •Reactive Monitoring:Alerts triggered after voltage thresholds are breached, which is too late. •Poor Selectivity:Frequent alerts for temporary, selfcorrecting events lead to unnecessary diagnostics. •Computationally Wasteful:Traditional systems reprocess raw data for every new question or query. The Vision: A Proactive, Future-Proof System •Fast & Scalable:Must be able to process massive data streams from smart meters, PMUs in real-time. •Efficient Storage & Computation:Enable long-term historical/temporal analysis without prohibitive costs. •High-Accuracy & Context-Aware:Leverage the grid's hierarchy to understand spatial correlations. Motivation Our key idea is to combine the contextual accuracy of hierarchical analysis with the performance benefits of data summarization to achieve high-accuracy, highperformance and low-cost analytics. 1. Model the Topology: We first map the grid's physical hierarchy, establishing the relationships between assets (meters → junction boxes → transformers). 2. Summarize Once: A one-pass summarization algorithm (e.g., Discrete Wavelet Transform) is applied to the time-series data to create a persistent, multi-scale summary that is orders of magnitude smaller than the raw data. 3. Run Anything: This compact summary becomes the single source for multiple analytics. It allows for dynamic, on-the-fly queries across any time-scale or spatiallevel without ever needing to re-process the raw data. Insights The HierarchyScope Framework Spatial (Hierarchical) Correlation •Smart meters are not independent; they are grouped by physical connections (e.g., distribution transformers). • A systemic issue (e.g., transformer fault) will manifest as a correlated pattern across a group. •An individual issue (e.g., faulty meter) will be an outlier within its group. Data Summarization • Efficiently produce compact representation of the data. • Energy/Resource efficient, fast/scalable, configurable. • Use-cases: monitoring, correlation analysis, etc. References A. J. van Rooij, V. Gulisano, and M. Papatriantafilou, “LoCoVolt: Distributed Detection of Broken Meters in Smart Grids through Stream Processing,” in Proceedings of the 12th ACM International Conference on Distributed and Event-based Systems, in DEBS ’18. New York, NY, USA: Association for Computing Machinery, June 2018, pp. 171–182. doi: 10.1145/3210284.3210298. B. W. R. Punter, O. Papapetrou, and M. Garofalakis, “OmniSketch: Efficient Multi-Dimensional High-Velocity Stream Analytics with Arbitrary Predicates,” Proc. VLDB Endow., vol. 17, no. 3, pp. 319–331, Nov. 2023, doi: 10.14778/3632093.3632098. C. I. Mytilinis, D. Tsoumakos, and N. Koziris, “Workload-aware wavelet synopses for sliding window aggregates,” Distrib Parallel Databases, vol. 39, no. 2, pp. 445–482, June 2021, doi: 10.1007/s10619-020-07307-w. Vinh Quang Ngo, Joris van Rooij and Marina Papatriantafilou Summaries Smart meters PMUs Data Summarization Query different levels of the hierarchy (e.g., child of some transformer_id, substation_id)? Query based on some augmented attributes (e.g., postal_code, customer_type, has_solar) ? Query using different time-granularity (e.g., group by 10m, 1h from last year, last 2 years). High-Voltage Transmission 230kV Distribution Transformers 240V Junction Boxes Household smart meters Junction Box A Junction Box A1 Junction Box A2 SM A11 SM A21 SM A22 Data Analytics Monitoring Substation Transformers 11kV Substation Transformers 69kV … … MSCA Doctoral Network RELAX-DN - EU Horizon 2021-27 -Relaxed Semantics Across the Data Analytics Stack Swedish Electricity Storage and Balancing Centre Göteborg Energi HierarchyScope: Spatio-Temporal Data Summarization for Efficient Voltage Forecasting and Anomaly Detection in Hierarchical Power Grids Poster