Electrical Conductor Vibration Detection for Noise-Robust Fault Sensing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting partial discharges in medium-voltage or high-voltage electrical devices suffer from high false detection and non-detection rates due to significant noise interference, particularly for low-amplitude discharges.
Innovation Solution
A method using machine learning to predict vibration signal patterns based on reference samples from normal operating conditions, allowing for accurate detection of abnormal events by calculating differences between measured and modeled signal values, with adaptive thresholds and alert mechanisms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If statistical processing of vibration measurements is used to detect partial discharges, then the detection method can be implemented, but the accuracy is limited due to significant noise interference
Solution Approach 1:
The patent applies preliminary action by acquiring a set of reference samples during a learning phase when no abnormal events are present, and using these samples to train a machine learning model before actual detection begins. This pre-processing of reference data enables the system to distinguish between normal noise and actual partial discharge signals during operation, thereby improving detection accuracy despite noise interference.
Solution Approach 2:
The patent creates a virtual reference model of normal operating conditions by copying and analyzing vibration patterns from the learning phase. This virtual model serves as a baseline for comparison during detection, allowing the system to identify deviations caused by partial discharges while filtering out typical noise, thus resolving the contradiction between detection capability and noise interference.
2Reliability
If a vibration measurement sensor is arranged on the insulator to monitor partial discharges, then local monitoring is enabled, but the false detection rate and non-detection rate remain relatively high
Solution Approach 1:
The patent implements feedback by continuously comparing current vibration measurements against the machine learning model trained on reference samples. The system uses the difference between actual and predicted values as feedback to identify abnormal events, adjusting detection thresholds based on learned normal variations. This feedback mechanism reduces both false detections and non-detections by dynamically adapting to the specific equipment's characteristics.
Solution Approach 2:
The patent applies parameter changes by transforming the detection approach from fixed statistical thresholds to adaptive parameters derived from machine learning. The system changes the detection parameters based on the specific characteristics of each electrical device learned during the reference phase, thereby improving reliability and reducing error rates compared to universal statistical methods.
3Measurement precision
If machine learning is used to create a prediction model for vibration signals, then detection accuracy is improved, but the complexity of the detection system increases
Solution Approach 1:
The patent reduces operational complexity by performing the complex machine learning model training in advance during a reference phase before deployment. Once the model is trained and stored, the actual detection process uses this pre-computed model for simple comparisons, thereby achieving high accuracy without requiring complex real-time computation during operation.
Solution Approach 2:
The system applies self-service by automatically learning and adapting to each specific electrical device's characteristics during the reference phase without requiring manual calibration or tuning. The machine learning model autonomously captures the unique vibration patterns of the equipment, eliminating the need for expert intervention in model customization and reducing overall system complexity.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly improves detection accuracy by distinguishing between nominal noise and abnormal events, reducing false positives and negatives, and adapting to the specific characteristics of each electrical device.
Implementation Method 1
measuring vibrations at the surface of an insulator around an electrical conductor
Data Source
AI summary
A method for detecting an abnormal event in an insulator of an electrical conductor of a medium-voltage or high-voltage electrical device. The method includes: acquiring a set of successive samples of a vibration signal associated with the electrical conductor; determining a modelled value of a following sample based on the acquired set of samples and on a prediction model, the prediction model being obtained through machine learning of the vibration signal based on a set of samples acquired in reference conditions in which the electrical conductor is free from any abnormal event; acquiring the following sample of the vibration signal; calculating a difference between the value of the acquired sample and the modelled value; and if the calculated difference is greater than a predetermined threshold, detecting an abnormal event in the electrical conductor.