Microgrid Fault Detection Using a Digital Twin Classifier
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electrical protection systems in microgrids, particularly those with inverter-based generators, struggle to detect and suppress short circuit faults due to low short circuit currents, and centralized protection systems are difficult to implement and require complex tuning.
Innovation Solution
A machine learning classifier system is trained using a digital representation of the microgrid to identify and locate electrical faults by simulating various fault scenarios, allowing for automatic configuration and reliable fault detection with minimal human intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If decentralized electrical protection systems based on local protection devices are used, then the system structure is simple, but the fault detection capability is insufficient due to low short circuit currents from inverter-based generators
Solution Approach 1:
The patent introduces a centralized controller as an intermediary that collects current measurements from multiple protection devices and performs advanced analysis. This mediator enables sophisticated fault detection algorithms to be implemented without requiring complex hardware modifications at each local protection device, thus improving detection capability while maintaining relatively simple device structure.
Solution Approach 2:
The centralized controller serves multiple functions: it acts as a data collection point for all protection devices, performs fault detection analysis, determines fault locations, and coordinates protection device operations. This multi-functionality consolidates complexity into a single unit while enabling sophisticated protection capabilities across the entire microgrid system.
2Measurement precision
If centralized protection schemes based on a central electronic controller are used, then the fault detection capability is improved, but the programming and fine-tuning complexity increases significantly
Solution Approach 1:
The system uses parameter changes in the form of weighted coefficients that are automatically adjusted based on the specific microgrid configuration. The centralized controller adapts its detection algorithms by modifying these parameters according to the actual system characteristics, reducing the need for manual fine-tuning while maintaining high detection accuracy.
Solution Approach 2:
The centralized controller automatically determines optimal protection settings and coordination parameters based on the digital representation of the microgrid and simulated fault data. This self-configuration capability reduces the need for manual programming and fine-tuning, allowing the system to adapt to different microgrid topologies and operating conditions automatically.
3Device complexity
If classical decentralized protection systems based on relays are used, then the device structure is simple, but the economic cost increases substantially and tuning becomes complex and lengthy
Solution Approach 1:
The system performs preliminary actions by creating a digital representation of the microgrid and simulating various fault scenarios before actual deployment. This pre-training phase allows the centralized controller to learn optimal detection parameters and coordination strategies in advance, eliminating the need for time-consuming on-site tuning and testing.
Solution Approach 2:
The patent creates a digital copy (digital representation) of the physical microgrid system that includes all components, connections, and operating parameters. This virtual model is used for training the protection system through simulated faults, allowing the system to be configured and tested virtually before deployment, thereby reducing actual commissioning time and costs.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method for training a machine learning classifier system for an electrical fault detection system comprises: acquiring a digital representation of a target microgrid electrical distribution system (2) comprising a plurality of electrical switchgear devices (10), power sources and loads, simulating electrical faults under different operating conditions and/or at a plurality of locations in the digital representation of the target electrical distribution system, determining the simulated electrical response of the electrical distribution system to each simulated electrical fault, generating classifier parameters by associating, by the classifier system, the determined simulated responses to the corresponding simulated electrical faults.