Dynamic Grid Island Event Localization at Component Level
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current predictive maintenance applications for complex networks, such as electrical grids, struggle to accurately identify the specific components causing interruptions, leading to inefficient response and control, especially during dynamic changes like islanding and remerging events.
Innovation Solution
A machine learning latent variable model is employed to disaggregate network events, allowing for the precise assignment of interruptions to individual components and predicting future behavior by using probabilistic modeling and iterative probability updates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current predictive maintenance applications are used, then interruption events can be assigned to a particular region of the grid, but the specific component causing the interruption cannot be identified
Solution Approach 1:
The patent segments the network into dynamically changing islands and further segments each island into individual components. The predictive maintenance application separately models island-level events and component-level events, allowing precise localization by combining both levels of analysis rather than treating the entire network as a single region.
Solution Approach 2:
The patent introduces an intermediary probabilistic model that connects island-level observations to component-level causes. This model acts as a mediator that translates regional event data into specific component assignments through probability propagation, enabling precise localization without requiring direct sensors on every component.
2Adaptability or versatility
If simple logic is used for predictive maintenance, then interruption events can be detected, but the logic becomes insufficient as network complexity increases
Solution Approach 1:
The patent makes the predictive maintenance logic dynamic by allowing the network topology to change over time through island formation and dissolution. The probabilistic models are updated iteratively as islands form and merge, enabling the system to adapt to changing network configurations rather than relying on static logic.
Solution Approach 2:
The patent changes the parameters of the predictive maintenance logic from fixed logical rules to probabilistic parameters that can be updated based on observed events. The model uses probability distributions for component failures and updates these parameters as new information becomes available, allowing the logic to adapt to network complexity.
3Productivity
If region-level event assignment is used, then the system can handle simple network configurations, but response efficiency decreases for complex networks with dynamic islanding
Solution Approach 1:
The patent adds another dimension to event assignment by operating at two levels simultaneously: island-level and component-level. This dimensional expansion allows the system to maintain region-level overview while also capturing component-specific details, preventing information loss and improving response efficiency for complex networks.
Solution Approach 2:
The patent implements feedback mechanisms where component-level probability assignments are updated based on island-level event observations, and vice versa. This bidirectional feedback ensures that information is not lost at either level but rather integrated to improve overall response efficiency and accuracy.
Data Source
AI summary
The present disclosure provides systems and methods that may advantageously apply machine learning to detect and ascribe network interruptions to specific components or nodes within the network. In an aspect, the present disclosure provides a computer-implemented method comprising: mapping a network comprising a plurality of islands that are capable of dynamically changing by splitting and/or merging of one or more islands, wherein the plurality of islands comprises a plurality of individual components; and detecting and localizing one or more local events at an individual component level as well as at an island level using a disaggregation model.