Grid Anomaly Detection Using FFT and Adaptive Signal Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting anomalies in industrial grids, such as voltage interruptions or drops, are costly and require expensive hardware and technical expertise, often leading to delayed or unnoticed failures due to decoupling from the drive system and reliance on fixed threshold values.
Innovation Solution
A method using Fast Fourier Transformation (FFT) and spectrogram analysis, combined with machine learning techniques like autoencoders and unsupervised learning models, to detect anomalies in grid data without additional equipment, enabling real-time, automated detection integrated with the powertrain system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If additional measurement sensors and network analyzers are used for grid analysis, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent creates a virtual copy of the grid signal processing capability by implementing FFT and spectrogram analysis software within the frequency converter's existing processor. This virtual instrument replicates the functionality of expensive physical network analyzers without requiring additional hardware, thereby maintaining measurement precision while eliminating the need for separate measurement devices
Solution Approach 2:
The frequency converter's existing processor and communication interfaces are made multi-functional by enabling them to perform both their original control functions and additional grid analysis functions. The same processor that controls the frequency converter also performs FFT transformations, spectrogram generation, and anomaly detection, eliminating the need for dedicated measurement hardware
2Ease of operation
If fixed threshold values are used for anomaly detection, then ease of operation is improved, but reliability deteriorates due to late or missed anomaly detection
Solution Approach 1:
The patent replaces static fixed threshold values with dynamic adaptive thresholds that automatically adjust based on learned normal operating patterns. The system continuously learns the grid's characteristic patterns through FFT and spectrogram analysis, then dynamically adapts detection thresholds to account for varying operating conditions, thereby improving reliability without sacrificing ease of operation
Solution Approach 2:
The system performs self-learning and self-adjustment by automatically analyzing grid signals, identifying normal patterns, and adapting detection parameters without requiring manual calibration or expert intervention. The frequency converter autonomously improves its anomaly detection capability by continuously learning from operational data
3Device complexity
If process signals from frequency converters are used for anomaly detection, then device complexity is reduced, but measurement precision deteriorates due to signal variations between machines
Solution Approach 1:
The patent transforms the raw process signals through FFT and spectrogram transformations, changing the parameter space from time-domain voltage/current measurements to frequency-domain representations. This parameter transformation makes the signals comparable across different machines by converting machine-specific temporal variations into standardized spectral features that can be universally analyzed
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Apparatus, System and Method for detecting anomalies in a grid are disclosed. The method comprising transforming data acquired from the grid (110, 110A-110C) based on at least one of a Fast Fourier Transformation (FFT) and a spectrogram of the data, wherein the data acquired comprises data associated with at least one of grid voltage, grid current, grid frequency, phase; fitting the data using a fitting function initialized using at least the transformed data, wherein the fitting function includes at least one of a sinusoidal function; or generating a lower representation of at least one of the data acquired and the transformed data; and detecting the anomaly in the grid (110, 110A-110C) based on at least one outlier detected in the fitted data or the lower representation of data using at least one of a parameter deviation and the similarity index.