IED Decision Logic Training for Fast, Reliable Fault Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating decision logic in protection and monitoring systems, such as those used in electric power systems, face challenges in achieving rapid and reliable fault detection without overfitting, particularly in complex scenarios, and require human expert knowledge for distinguishing training cases.
Innovation Solution
The system employs an iterative procedure to train machine learning models using a training kernel technique, where weighting functions are adjusted automatically to optimize decision-making speed and accuracy, allowing the models to process time-series inputs and generate outputs indicative of corrective actions without human expert classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If machine learning models are trained to detect faults quickly, then decision speed is improved, but the models may overfit and reduce reliability
Solution Approach 1:
The patent applies dynamics by making the training kernel adjustable and adaptive during the training process. The training kernel is modified based on the complexity of training cases, allowing the model to dynamically adapt its learning behavior. This enables the model to learn quickly from simple cases while spending more time on complex cases, thereby improving decision speed without sacrificing reliability through overfitting.
Solution Approach 2:
The patent changes the parameter of the training kernel during the training process. By modifying the training kernel based on training case complexity, the model can adjust its learning rate and focus. This parameter change allows the system to optimize both speed and reliability by allocating different amounts of learning effort to different types of training cases.
2Reliability
If human experts classify training cases to improve accuracy, then dependability is improved, but the process complexity increases
Solution Approach 1:
The patent applies self-service by enabling the machine learning model to automatically classify and prioritize training cases based on their complexity, without requiring human expert intervention. The model itself determines which cases are simple and which are complex, and adjusts its training kernel accordingly. This self-service approach maintains dependability while eliminating the complexity associated with manual expert classification.
Solution Approach 2:
The patent replaces the mechanical process of human expert classification with an automated computational system. Instead of relying on human experts to manually classify training cases, the system uses algorithmic methods to automatically assess case complexity and adjust training parameters. This substitution maintains or improves dependability while significantly reducing process complexity.
3Manufacturing precision
If more training cases are included to improve accuracy, then manufacturing precision is improved, but training time increases
Solution Approach 1:
The patent applies partial action by focusing training effort on the most critical aspects of fault detection. Instead of treating all training cases equally, the system identifies and prioritizes complex cases that require more attention, while using simpler cases more efficiently. This allows the model to achieve high accuracy by concentrating computational resources where they are most needed, rather than uniformly processing all training data.
Solution Approach 2:
The patent uses dynamics to adapt the training process based on case complexity. By modifying the training kernel during training, the system can quickly process simple cases and allocate more time to complex cases. This dynamic approach enables the model to achieve high detection accuracy across diverse training cases while minimizing overall training time through efficient resource allocation.
Data Source
AI summary
To generate a decision logic for an IED, at least one machine learning model is trained in an iterative machine learning model training. Weighting functions are used to weight samples in the iterative machine learning model training. Weighting function(s) associated with one or several training cases are automatically modified in the iterative machine learning model training.