Microgrid Inverter IGBT Aging Prediction for Reliable Junction Temperature

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The reliability evaluation of microgrid inverters, particularly IGBTs, is challenged by aging-related inaccuracies in electrothermal coupling models due to fatigue damage, leading to inaccurate junction temperature measurements and reduced monitoring capabilities.

Innovation Solution

A fusion reliability evaluation method and system based on segmented long short-term memory (LSTM) is introduced, combining IGBT reliability analysis with aging monitoring and machine learning algorithms to predict the aging process, update electrothermal coupling model parameters, and calculate real-time thermal stress and cumulative damage.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If indirect measurement using pre-constructed electrothermal coupling model is used, then online monitoring capability is improved, but measurement precision deteriorates due to model aging

Engineering Contradiction:

Improveonline monitoring capabilityVSAvoidjunction temperature measurement accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The system performs preliminary aging prediction using segmented LSTM to forecast future aging states before they occur. This allows the electrothermal coupling model to be proactively updated with predicted aging parameters, ensuring the model remains accurate without waiting for actual aging to degrade performance. The preliminary prediction of aging trends enables preventive model correction.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements a feedback mechanism where segmented LSTM continuously predicts aging based on historical data, and these predictions feed back to update the electrothermal coupling model parameters. This closed-loop feedback ensures the model adapts to actual aging conditions, maintaining measurement precision while preserving online monitoring capabilities.

Inventive Principle:

Principle #23Feedback

2Measurement precision

If direct measurement with integrated sensor is used, then measurement precision is improved, but device complexity increases

Engineering Contradiction:

Improvejunction temperature measurement accuracyVSAvoidIGBT module structure complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces segmented LSTM as an intermediary computational model that bridges the gap between easily obtainable electrical parameters and the difficult-to-measure junction temperature. Instead of directly measuring temperature with complex sensors, the LSTM model indirectly infers temperature through electrical characteristics, simplifying the overall system while maintaining accuracy.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces the physical sensor-based mechanical measurement system with a data-driven computational model (segmented LSTM). This substitution eliminates the need for physical integrated sensors, reducing device complexity while maintaining measurement capability through algorithmic prediction based on electrical parameters.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Reliability

If machine learning algorithm is introduced for aging prediction, then reliability evaluation accuracy is improved, but calculation efficiency decreases

Engineering Contradiction:

Improvereliability evaluation accuracyVSAvoidcalculation efficiency

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent segments the LSTM prediction process into distinct phases: training phase with historical data, and inference phase with real-time data. This segmentation allows computationally intensive training to be performed offline when calculation resources are abundant, while real-time prediction uses the trained model for efficient online operation, thus balancing accuracy and efficiency.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The segmented LSTM model performs preliminary processing and feature extraction during the training phase, preparing optimized prediction structures in advance. This preliminary action reduces the computational burden during real-time reliability evaluation, improving calculation efficiency without sacrificing prediction accuracy.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS11913986B2Reliability evaluation method and system of microgrid inverter IGBT based on segmented LSTM

Publication Date: 2024.02.27 WUHAN UNIV

AI summary

A reliability evaluation method and system for a microgrid inverter IGBT based on segmented long short-term memory (LSTM) is disclosed, including steps as follows. An electrothermal coupling model is constructed to obtain real-time junction temperature data. The original LSTM algorithm is improved to obtain a segmented LSTM prediction network for the aging characteristics of the IGBT. The monitoring value of the IGBT aging parameter is used to perform segmented LSTM prediction to obtain the predicted aging process, and the threshold values of different aging stages are categorized. An aging correction is performed on the aging parameter of the electrothermal coupling model to ensure the accuracy of the junction temperature data. Rainflow-counting algorithm is used to calculate real-time thermal stress load distribution of the IGBT. The fatigue damage theory and the Lesit life prediction model are combined to calculate the real-time cumulative damage and predicted life of the IGBT.