Solar Panel Network Fault Localization with CNN Models

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional fault detection and diagnosis techniques for solar panel networks are limited in identifying faults at a broader level or studying one specific fault type at a time, leading to significant energy loss and potential hazards, and lack the ability to accurately detect and localize various fault types within the network.

Innovation Solution

A processor-implemented method and system using a fault detection, diagnosis, and localization (FDDL) model trained with a plurality of fault and no-fault datasets, utilizing a convolutional neural network (CNN) to identify and locate faults in solar panel networks by simulating various fault scenarios and environmental conditions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional fault detection and diagnosis techniques are used, then the system is simpler to implement, but the fault detection precision and localization capability are insufficient

Engineering Contradiction:
Improvefault detection precisionVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent creates a digital twin (virtual model) of the solar panel network that replicates the physical system's structure and behavior. This virtual model is trained with simulated fault data to learn fault patterns, enabling high-precision fault detection and localization without requiring complex physical sensing infrastructure at every panel. The digital twin serves as a virtual copy that performs the complex analysis work.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The system performs preliminary training of the digital twin model using extensively simulated fault scenarios before actual deployment. During this offline preparation phase, the model learns to recognize various fault types and their signatures. When deployed, the pre-trained model can quickly detect and diagnose faults in real-time without requiring complex real-time computation, thus achieving high precision without proportional increase in operational system complexity.

Inventive Principle:
Principle #10Preliminary action

2Loss of energy

If conventional FDD techniques are used, then the implementation cost is lower, but energy loss due to undetected faults increases

Engineering Contradiction:
Improveenergy lossVSAvoidFDD system complexity
Core Design Contradiction:
Loss of energyVSDevice complexity

Solution Approach 1:

By using a digital twin that replicates the solar panel network, the system can simulate and detect faults virtually without interfering with actual power generation. The digital twin processes data to identify faults that would otherwise cause energy loss, enabling early intervention while maintaining simple physical infrastructure.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The system implements continuous monitoring where the digital twin receives real-time operational data from the solar panel network, compares actual performance against expected performance, and provides feedback about detected faults. This closed-loop feedback mechanism enables proactive fault detection and response, minimizing energy loss from undetected faults while maintaining a relatively simple system architecture.

Inventive Principle:
Principle #23Feedback

3Adaptability or versatility

If conventional FDD techniques are used, then the system architecture is simpler, but the capability to detect multiple fault types simultaneously is limited

Engineering Contradiction:
Improvemulti-fault detection capabilityVSAvoidFDD system complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The digital twin is designed as a universal detection system that can identify multiple types of faults (hot spot faults, arc faults, line-line faults, line-ground faults, etc.) using a single integrated model. Rather than requiring separate detection systems for each fault type, the multi-functional digital twin analyzes operational data to detect various fault types simultaneously, reducing overall system complexity while enhancing versatility.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system detects different fault types by analyzing changes in electrical parameters (current, voltage, power) and their patterns. The digital twin is trained to recognize distinct parameter change signatures associated with different fault types, enabling multi-fault detection through parameter analysis rather than requiring separate physical sensors for each fault type.

Inventive Principle:
Principle #35Parameter changes

4Measurement precision

If conventional FDD techniques are used, then the response time for fault identification is faster, but the localization accuracy is poorer

Engineering Contradiction:
Improvefault localization accuracyVSAvoidfault identification time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The digital twin replicates the exact physical layout and electrical characteristics of the solar panel network, enabling precise spatial localization of faults. By comparing the virtual model's expected behavior with actual measurements, the system can pinpoint fault locations accurately without requiring time-consuming physical inspection or complex signal propagation analysis.

Inventive Principle:
Principle #26Copying

Data Source

PatentEP3809592B1Methods and systems for fault detection, diagnosis and localization in solar panel network
Publication Date: 2025.07.16 TATA CONSULTANCY SERVICES LTD
  • EP3809592B1 patent drawingFigure 1
  • EP3809592B1 patent drawingFigure 2
  • EP3809592B1 patent drawingFigure 3

AI summary

This disclosure relates generally to the methods and systems for fault detection, diagnosis and localization in solar panel network. Conventional fault detection and diagnosis (FDD) techniques for the solar panel network are limited and confined to identifying faults either at voltage level or current level, or to studying one specific fault type at a time. The present disclosure solve the problems of detecting various fault types present inside the solar panel network and identifying associated fault locations, by generating a fault detection, diagnosis and localization (FDDL) model. The convolutional neural network (CNN) model is trained with fault datasets and no-fault datasets covering various fault scenarios and no-fault scenarios respectively, to generate the FDDL model. The plurality of fault datasets and the plurality of no-fault datasets are determined based on the network simulation model of the solar panel network.