Transformer Failure Prediction Using Variable Worth Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The aging electric grid in the United States faces challenges in predicting transformer failures, leading to inefficiencies and potential catastrophic failures, due to the high cost and difficulty of frequent site visits for maintenance, especially with transformers from the 1960s and 1970s still in use.
Innovation Solution
A computer-based system that predicts the probability of electrical transformer failure by analyzing historical data, selecting the most relevant variables, training models, and validating their fit to historical data to accurately forecast transformer failures, allowing for targeted maintenance and reducing emissions by replacing or repairing transformers before they fail.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If frequent periodic site visits by maintenance personnel are conducted to determine transformer longevity, then the reliability of transformer monitoring is improved, but the cost and difficulty of implementation increase
Solution Approach 1:
The transformer monitoring system enables self-service by allowing transformers to automatically report their own operational status, temperature, and health metrics through embedded sensors and communication modules. This eliminates the need for manual site visits while maintaining continuous monitoring reliability, directly resolving the contradiction between monitoring reliability and operational complexity.
Solution Approach 2:
The patent replaces the mechanical system of manual site visits with an automated electronic monitoring system using sensors, data collectors, and communication networks. This substitution maintains reliable monitoring while dramatically reducing the operational complexity and cost associated with frequent physical inspections.
2Productivity
If remote monitoring equipment is installed on all transformers to enable continuous tracking, then the productivity of maintenance operations is improved, but the installation cost becomes prohibitive
Solution Approach 1:
The system applies local quality by strategically deploying monitoring equipment only on transformers that are predicted to be at high risk of failure, rather than uniformly installing equipment on all transformers. This targeted approach improves maintenance productivity while significantly reducing the total quantity of monitoring equipment required, making the solution economically viable.
Solution Approach 2:
The patent uses parameter changes by analyzing variations in operational parameters (temperature, load, vibration) to predict transformer failure risk. This allows the system to dynamically identify which transformers need monitoring without requiring universal equipment installation, thereby improving productivity while controlling equipment quantity and cost.
3Loss of time
If transformers from the 1960s and 1970s continue to be used in the electric grid, then the loss of time for replacement is reduced, but the risk of catastrophic failure increases
Solution Approach 1:
The system applies preliminary action by continuously monitoring transformer health parameters and predicting failures before they occur. This allows maintenance teams to proactively replace aging transformers from the 1960s and 1970s before catastrophic failure, reducing both the risk of failure and the time loss associated with emergency replacements by planning ahead.
Solution Approach 2:
The patent implements feedback by continuously collecting operational data from transformers, analyzing trends, and providing real-time information about transformer health status. This feedback loop enables early detection of deterioration in aging transformers, allowing timely intervention that maintains reliability while minimizing unplanned downtime and replacement time loss.
Data Source
AI summary
A computing device predicts a probability of a transformer failure. An analysis type indicator defined by a user is received. A worth value for each of a plurality of variables is computed. Highest worth variables from the plurality of variables are selected based on the computed worth values. A number of variables of the highest worth variables is limited to a predetermined number based on the received analysis type indicator. A first model and a second model are also selected based on the received analysis type indicator. Historical electrical system data is partitioned into a training dataset and a validation dataset that are used to train and validate, respectively, the first model and the second model. A probability of failure model is selected as the first model or the second model based on a comparison between a fit of each model.