Wavelet Transform Fault Detection in Doubly Fed Induction Generators

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current fault monitoring methods for double-fed asynchronous machines (DFIDs) in wind turbines are inadequate, as they often require machine shutdown for electrical measurements, are unreliable due to averaging over time, and struggle to distinguish between real and fictitious anomalies, especially with frequency variations caused by changing wind speeds.

Innovation Solution

A method involving real-time computer-based monitoring using current and voltage measurements from the stator, combined with relative rotational speed, performing a wavelet transform on stator power to generate a harmonic representation, and comparing it to a reference to detect faults, with the option to include reactive power and stator voltage in the reference selection.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If Fourier transform harmonic analysis is used to detect faults, then characteristic frequencies can be identified, but the method averages measurements over time which prohibits fine monitoring of variations and reduces reliability when frequency varies

Engineering Contradiction:
Improvefault detection precisionVSAvoidanomaly detection reliability
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent changes the mathematical transformation parameter from Fourier transform to wavelet transform. This allows the analysis to adapt to varying frequencies in real-time, maintaining both measurement precision and reliability by capturing transient features without time-averaging that obscures variations.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The wavelet transform provides a dynamic time-frequency analysis that adapts to changing operating conditions. Unlike static Fourier analysis, it can track frequency variations over time, making the fault detection system responsive to dynamic changes in MADA operation while maintaining reliability.

Inventive Principle:
Principle #15Dynamics

2Productivity

If vibration analysis is used to monitor MADA during operation, then energy production continues uninterrupted, but the method is unreliable and only detects serious defects requiring emergency measures

Engineering Contradiction:
Improveenergy production continuityVSAvoidfault detection reliability
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent replaces mechanical vibration analysis with electrical signal analysis using wavelet transform. This substitution maintains the ability to monitor during operation (preserving productivity) while significantly improving reliability by detecting faults at earlier stages through more sensitive electrical measurements.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The wavelet transform acts as an intermediary that processes electrical measurements to extract fault information with high reliability. It bridges the gap between continuous operation monitoring and accurate fault detection, enabling early warning before serious damage occurs.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Measurement precision

If characteristic frequency analysis is performed on electrical currents, then fault frequencies can be identified, but the method becomes complex and unreliable when frequencies are close to fundamental frequency and difficult to distinguish

Engineering Contradiction:
Improvefrequency identification precisionVSAvoidmonitoring method complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent moves from analyzing only frequency domain to joint time-frequency domain analysis using wavelet transform. This adds a temporal dimension that separates closely spaced frequencies by their time-localized characteristics, reducing complexity while improving precision in distinguishing fault frequencies near the fundamental frequency.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

4Measurement precision

If electrical measurements are performed with MADA stopped, then accurate measurements can be obtained, but production is interrupted and energy generation is lost

Engineering Contradiction:
Improveelectrical measurement precisionVSAvoidenergy production
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The wavelet transform enables dynamic signal analysis during operation, allowing precise fault detection without stopping the MADA. The time-frequency localization capability maintains measurement precision even under varying operating conditions, eliminating the need to interrupt production for measurements.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3650876B1Detection of electrical fault in a generator
Publication Date: 2021.05.12 ELECTRICITE DE FRANCE
  • EP3650876B1 patent drawingFigure 1~2
  • EP3650876B1 patent drawingFigure 3~4
  • EP3650876B1 patent drawingFigure 5

AI summary

A method for monitoring and detecting faults in a doubly fed induction machine comprising a stator and a rotor, said method being implemented by computer means. The method comprises: a. receiving (101) a measurement of the stator current and voltage, and a measurement of the relative rotational speed between the stator and the rotor; b. deducing (102) a stator power as a function of time; c. performing (105) a harmonic analysis of the obtained stator power, including applying a wavelet transform to the stator power; d. comparing (106) the obtained harmonic representation to a reference harmonic representation, said reference harmonic representation being selected based on the rotational speed measurement, so as to detect faults in said machine based on any differences identified by the comparison.