In-Field Fault Detection Using Neural Networks

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional power supply systems face challenges in reliably and efficiently detecting fault types in power transmission lines due to resource-intensive manual efforts and limitations in adapting to new data patterns, requiring substantial computational and storage resources.

Innovation Solution

An in-field detection apparatus employing a preprocessing unit for normalizing voltage and current data, coupled with an artificial neural network trained on both simulated and historical fault records, which maps data to fault type scores for real-time detection and classification, utilizing a feed-forward, convolutional, or recurrent neural network architecture with minimal resource requirements.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If conventional rule-based fault classification is used, then fault detection can be performed, but substantial manual expert-driven effort is required for adding new rules and improvements

Engineering Contradiction:
Improveadaptability to new fault typesVSAvoidmanual expert-driven effort
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent replaces the mechanical rule-based system with an artificial neural network that automatically learns fault patterns from data. The neural network substitutes manual rule creation with automated machine learning, where the system adapts to new fault types through training on historical data rather than requiring manual expert intervention to create new rules.

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

Solution Approach 2:

The neural network performs self-learning and self-improvement by automatically training on fault records and extracting patterns without continuous manual intervention. The system serves itself by autonomously adapting to new fault types through ongoing training processes, reducing the need for external expert input.

Inventive Principle:
Principle #25Self-service

2Measurement precision

If extensive signal processing and custom rules are applied, then fault classification can be achieved, but computational resources and storage requirements increase substantially

Engineering Contradiction:
Improvefault classification accuracyVSAvoidcomputational resources
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent applies preliminary normalization to the input signals before processing, which simplifies subsequent analysis. By pre-processing the data to remove variations in magnitude and scale, the neural network requires less computational effort during classification while maintaining high accuracy. This preliminary action reduces the complexity of the main processing task.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If manual rule-based classification is used, then existing fault types can be detected, but the system cannot effectively adapt to new previously unseen fault patterns

Engineering Contradiction:
Improvedetection reliabilityVSAvoidcapability to detect new fault types
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent implements a dynamic system where the neural network can be retrained with new fault data to adapt to changing conditions. Unlike static rule-based systems, the neural network's weights and parameters can be updated through continuous learning from new fault records, enabling the system to maintain high detection reliability for both known and emerging fault types.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3460494B1A method and apparatus for automatic detection of a fault type
Publication Date: 2023.01.25 SIEMENS AG
  • EP3460494B1 patent drawingFigure 1~2

AI summary

An in-field detection apparatus and method for automatic detection of a fault type of a fault having occurred at power transmission lines of a power supply system. The in-field detection apparatus (1) comprises: a preprocessing unit (2) adapted to normalize measured voltage raw data and/or current raw data of the power transmission lines; and a processing unit (3) configured to execute an artificial neural network stored in a local memory of said in-field detection apparatus (1), wherein the normalized voltage and/or current raw data is mapped to fault type scores of different predefined fault types and wherein the fault type scores are evaluated by said processing unit to detect the fault type of the occurred fault.