Smart Meter Anomaly Detection Using Edge Filtering and Shared Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing smart meter systems face challenges in detecting anomalies such as leaks or fraud without increasing complexity or data exchange volume, which is critical for battery-powered meters that require long service life.
Innovation Solution
A method involving Edge Computing where each meter acquires and preprocesses primary measurements, selecting relevant data for transmission to a remote processing device, which dynamically adjusts detection thresholds based on anomaly rates to enhance detection precision and robustness while minimizing network traffic and power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If meters perform local preprocessing and selective transmission of measurements, then data exchange volume is reduced and power consumption decreases, but detection precision may be compromised without sufficient statistical evaluation
Solution Approach 1:
The system divides the anomaly detection process into two segments: local preprocessing at the meter level (selecting relevant measurements from primary measurements) and centralized statistical evaluation at the remote processing device level. This segmentation allows energy-efficient local filtering while maintaining detection precision through remote statistical analysis of aggregated data from multiple meters.
Solution Approach 2:
The remote processing device acts as an intermediary that receives selected measurements from multiple meters, performs statistical evaluation to dynamically adjust detection thresholds, and feeds back adjusted thresholds to meters. This intermediary role enables centralized intelligence to enhance detection precision without requiring continuous full-data transmission from energy-constrained meters.
2Device complexity
If detection thresholds are statically configured in meters, then device complexity is reduced, but detection robustness deteriorates without adaptive adjustment capability
Solution Approach 1:
Detection thresholds are pre-configured in meters for initial operation, enabling simple local preprocessing. The remote processing device then performs preliminary statistical evaluation on aggregated measurements and dynamically adjusts thresholds, which are subsequently uploaded to meters. This preliminary action approach maintains low meter complexity while progressively enhancing detection robustness through remote adaptation.
Solution Approach 2:
The system transitions from static threshold configuration to dynamic threshold adjustment. The remote processing device continuously evaluates anomaly rates from multiple meters and adaptively modifies detection thresholds based on observed patterns. This dynamic adaptation enhances detection robustness against evolving anomaly types while maintaining simple meter architecture through centralized intelligence.
3Reliability
If all primary measurements are transmitted regularly to the remote device, then detection robustness is improved through comprehensive data analysis, but network traffic increases and battery life is reduced
Solution Approach 1:
The meter extracts and transmits only the most relevant measurements that meet selection criteria, rather than transmitting all primary measurements. This extraction process reduces network traffic and battery consumption while providing the remote processing device with sufficient data for robust anomaly detection through statistical evaluation of selected measurements from multiple meters.
Solution Approach 2:
Instead of transmitting complete measurement datasets, the system transmits a partial set of selected measurements that are most indicative of anomalies. This partial action approach reduces energy consumption and network traffic while maintaining detection robustness, as the remote device compensates by aggregating and statistically evaluating measurements from multiple meters.
4Measurement precision
If meters are equipped with advanced anomaly detection algorithms, then detection precision is improved, but device complexity and reprogramming requirements increase
Solution Approach 1:
Instead of embedding complex anomaly detection algorithms in meters, the system inverts the approach by placing sophisticated statistical evaluation and threshold adjustment algorithms in the remote processing device. Meters perform only simple preprocessing and transmission of selected measurements, while the remote device handles the computationally intensive anomaly detection logic, thereby improving detection precision without increasing meter complexity.
Solution Approach 2:
The remote processing device autonomously performs statistical evaluation of aggregated measurements and automatically adjusts detection thresholds without requiring manual reprogramming of meters. This self-service capability enables the system to adapt to changing anomaly patterns while maintaining simple meter architecture, as the intelligence resides in the remote device that serves the entire meter network.
Data Source
Figure 1
AI summary
Method for monitoring a set of meters (1, 2, 3, 4) which are connected to the same remote processing device (6), comprising the first steps, implemented in each meter, of: - acquiring primary measurements of a quantity representative of the occurrence of an anomaly; - selecting, among the primary measurements, relevant measurements which satisfy a relevance criterion; - regularly transmitting the relevant measurements to the remote processing device; and the second steps, implemented in the remote processing device (6), of: - for each meter, performing a comparison of the relevant measurements with a detection threshold common to all the meters in the set of meters, in an attempt to detect an anomaly associated with said meter; - dynamically adjusting the detection threshold according to a detected rate of anomalies.