Anomaly Detection in Electrical Devices Using Power Data Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to effectively detect and diagnose intermittent and anomalous behavior in electrical devices, particularly in real-time, due to limitations in power quality monitoring and reliance on specific threshold values, which can lead to financial losses and unreliable product performance.
Innovation Solution
A real-time monitoring and diagnostic system that uses a clustered-based pattern matching algorithm to detect anomalies in energy consumption data, including current, voltage, and power consumption, and generates anomaly detection rules based on baseline data, allowing for identification of internal states and anomaly detection without requiring pre-defined threshold values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If total switching time is used to determine diagnostic condition, then power electronic device monitoring is simplified, but intermittent behavior occurring for small duration cannot be identified
Solution Approach 1:
The patent segments the monitoring approach by dividing power data into multiple parameters (power consumption, current, voltage, power factor) and further segmenting time into different intervals (first time period, second time period). This allows detection of intermittent anomalies that occur during specific time periods while maintaining manageable system complexity through structured data organization.
Solution Approach 2:
The patent transitions from single-dimensional switching time monitoring to multi-dimensional power data analysis by incorporating power consumption, current, voltage, and power factor measurements across different time periods. This dimensional expansion enables precise detection of intermittent behaviors that would be invisible in single-parameter monitoring.
2Measurement precision
If threshold reference value is provided as input for anomaly detection, then anomaly detection accuracy is improved, but additional appliance specific information is required which may not be available
Solution Approach 1:
The system performs self-service by automatically establishing baseline power consumption patterns through learning during a first time period, then using this self-generated baseline for anomaly detection during a second time period. This eliminates the need for external threshold reference values or appliance-specific information input, as the system creates its own reference standards from operational data.
Solution Approach 2:
The patent implements preliminary action by collecting and analyzing power data during an initial first time period to establish baseline patterns before actual anomaly detection begins. This preliminary learning phase prepares the system with necessary reference information, enabling accurate anomaly detection without requiring pre-configured thresholds or external knowledge.
3Adaptability or versatility
If time series data analysis is used for anomaly detection, then usage pattern variations are captured, but reliability decreases when usage pattern changes due to demand expectation
Solution Approach 1:
The patent applies dynamics by implementing a two-phase approach where the system adapts to usage patterns during a first time period, then transitions to anomaly detection mode during a second time period. This dynamic phase transition allows the system to learn and adapt to varying usage patterns while maintaining reliable anomaly detection through consistent comparison against established baselines, rather than relying on static thresholds.
Data Source
AI summary
A system and method configured in an electrical device for providing a monitoring solution that identifies the intermittent or anomalous behavior of electronic/electric appliances when in use, comprising: receiving present real-time power features like current, voltage, power factor, power consumption of the associated device, and maintaining a history of past real-time readings of the device in its ideal working periods; forming rule-based clusters of different internal states of the device from the past history data at ideal working periods; determining which of said present readings are correlated; computing a deviation between at least some of said present and at least some of said clusters; and declaring an anomaly when said present readings deviation exceeds a predetermined threshold from all of the said clusters.