Power Asset Condition Classification with Missing Data Replacement

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for condition classification of power network assets face challenges in accurately analyzing assets when not all required parameter values are available, leading to unreliable results due to missing data.

Innovation Solution

Implementing a method that combines automatic classification with a missing data replacement procedure to generate substitute values for unavailable parameters, ensuring reliable condition classification even when not all input values are present.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a tool processes a large number of parameter values for automatic classification, then the accuracy of condition analysis is improved when all parameters are available, but the tool becomes incapable of analyzing assets when not all required parameter values are available

Engineering Contradiction:

Improvecondition analysis accuracyVSAvoidcapability to handle missing data

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system performs preliminary actions by training multiple classification models in advance, each specialized for different subsets of parameters. This allows the system to quickly select an appropriate pre-trained model when data is missing, rather than attempting to process incomplete data with a single comprehensive model.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The classification task is segmented into multiple specialized models, each handling specific parameter combinations. This segmentation allows the system to match the available data subset with the most appropriate model, improving both accuracy and adaptability to missing data scenarios.

Inventive Principle:

Principle #1Segmentation

2Reliability

If historical data is used to train a tool for processing multiple parameter values, then the training accuracy is improved, but the training process becomes challenging when historical data contains incomplete parameter values for most assets

Engineering Contradiction:

Improvetraining accuracyVSAvoidtraining process complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The training process is segmented into multiple independent model training tasks. Each model is trained on a specific subset of historical data that contains complete values for its required parameters, eliminating the complexity of handling incomplete data in a single training process.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system trains multiple models with varying degrees of parameter coverage, including some models trained on subsets of parameters. This partial action approach ensures that even if not all parameters are available, there exists a model trained specifically for that parameter subset, maintaining training reliability.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20260051732A1Methods and devices for condition classification of power network assets

Publication Date: 2026.02.19 HITACHI ENERGY LTD

AI summary

Methods and devices for a condition classification of a power network asset of a power network asset are disclosed. The methods and devices may combine an automatic classification procedure with a missing data replacement procedure.