Anomaly Detection in High-Resolution Electrical Measurement Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Renewable energy sources like solar farms face challenges in consistent operation due to irregular behavior, leading to significant plant failure rates, reduced equipment life, and unplanned outages, primarily because of the difficulty in processing and detecting anomalous behavior and faults in high-resolution electrical measurement data.
Innovation Solution
The integration of micro-synchrophasor measurement units (μPMU) and power quality monitors (PQM) with machine learning techniques, such as CLARANS clustering, for remote and automatic anomaly detection, utilizing a grid data unit that applies synchronized timestamps to high-resolution, time-stamped data for efficient processing and storage in a time-series database.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution electrical measurement data is collected from sensors, then measurement precision is improved, but data processing complexity increases
Solution Approach 1:
The patent segments high-resolution electrical measurement data into multiple feature sets including voltage, current, frequency, power, and other electrical parameters. Each feature set is processed independently through machine learning algorithms, transforming the complex high-resolution data into manageable components for anomaly detection.
Solution Approach 2:
The patent replaces traditional mechanical data processing methods with machine learning techniques. Neural networks and other ML algorithms automatically analyze the segmented electrical measurement data, substituting manual or rule-based processing with adaptive computational models that handle high-resolution data more efficiently.
2Measurement precision
If machine learning techniques are used for anomaly detection, then detection accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent divides the electrical measurement data into distinct feature sets (voltage, current, frequency, power, etc.) before applying machine learning algorithms. This segmentation allows multiple simpler models to process different aspects of the data simultaneously, achieving high detection accuracy while distributing computational complexity across multiple specialized algorithms rather than one complex model.
3Measurement precision
If high volume measurement data is processed, then anomaly detection capability is improved, but processing time increases
Solution Approach 1:
The patent segments high-volume electrical measurement data into multiple parallel feature sets that can be processed simultaneously through different machine learning algorithms. This parallel processing approach maintains comprehensive anomaly detection capability while reducing overall processing time compared to sequential analysis of the complete data set.
Data Source
AI summary
A method of processing high resolution electrical measurement data is disclosed including obtaining high resolution electrical measurement data related to time series data of a parameter measured from an electrical power grid system. The time series data comprises a first data points set transformed to feature vector format data where the time series data is grouped into a plurality of datasets, each dataset representing a subset of the first data points set. A statistical data clustering scheme is performed to generate distinct cluster patterns from the feature vector format data comprising a first cluster relating to a first electrical trend, a second cluster relating to a second, different electrical trend, and an outlier data pattern that is part of the first or second cluster. The outlier data pattern is far from its respective cluster centre. An anomalous event detection is based at least in part on the outlier data.