Grid Voltage Waveform Anomaly Classification Using Spectrograms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Utility grids face challenges in detecting and classifying anomalies in electrical waveforms, which can lead to equipment malfunctions and power outages, as existing systems lack the capability to identify deviations at high time resolutions and predict impending faults effectively.
Innovation Solution
A system utilizing high-resolution measurement devices to capture electrical waveforms, analyze them for anomalies, and employ machine learning models to classify and predict anomalies such as sag, swell, interruption, flicker, oscillatory transients, and harmonics, generating a signature identifier for targeted actions to prevent outages.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution measurement devices are used to capture electrical waveforms, then anomaly detection precision is improved, but device complexity increases
Solution Approach 1:
The system segments the complex anomaly detection task into multiple processing stages: waveform capture by measurement devices, preprocessing to generate features, spectrogram generation, and machine learning-based classification. This segmentation allows high-resolution measurement without requiring a single overly complex device, distributing complexity across modular components.
Solution Approach 2:
The patent introduces an intermediary processing layer that transforms raw high-resolution waveform data into spectrogram representations. This intermediary step simplifies the data structure, making it more amenable to machine learning analysis while preserving the high-resolution information needed for precise anomaly detection.
2Measurement precision
If machine learning models are employed to classify anomalies, then classification accuracy is improved, but computational requirements increase
Solution Approach 1:
The system performs preliminary processing of waveform data by generating spectrograms and extracting features before applying machine learning models. This preliminary action reduces the dimensionality and complexity of the input data, allowing machine learning models to achieve high classification accuracy with reduced computational energy consumption compared to processing raw high-resolution waveforms directly.
3Reliability
If continuous monitoring at high sampling rates is implemented, then reliability of anomaly detection is improved, but loss of energy increases
Solution Approach 1:
The system implements periodic monitoring with variable sampling rates rather than continuous high-rate sampling. The measurement devices capture waveforms at high sampling rates when anomalies are detected or suspected, and at lower rates during normal operation. This periodic approach maintains detection reliability while significantly reducing overall energy consumption compared to continuous high-rate monitoring.
Data Source
AI summary
Anomaly detection, classification, and prediction is provided. A system can include one or more processors coupled with memory. The system can identify voltage waveform data corresponding to electricity distributed over a utility grid and measured by a metering device. The system can detect, based on a comparison with baseline voltage waveform data, an anomaly in at least a portion of the voltage waveform data. The system can generate spectrogram data for the at least the portion of the voltage waveform data comprising the anomaly. The system can determine, via a model trained with machine learning, a type of the anomaly based on the spectrogram data. The system can provide an indication of the type of the anomaly to cause an action to be performed on the utility grid responsive to determination of the type of anomaly.