Transformer health indexing (THI) is an industrial strategy to rank transformers within a fleet according to their intervention priority. A THI score can help in ensuring commendable decisions on repair, replacement, and refurbishment of transformers in a grid. They can also determine the risk, reliability or both for each asset simultaneously. Nowadays, THI strategies are at the core of many “wholesome” but elaborate and expensive asset performance management (APM) suites. From the client’s perspective the only daunting question is, “how long do we have before this one goes out?”. An excellent discussion on the use of Ronin AI to ensure production uptime and optimal transformer performance for steel makers is also available.
With the advent of AI, Seetalabs’ has developed a tool to predict the THI score in real time using typical oil condition monitoring data from transformers. But the real question is, what is the best way to use Ronin AI to ensure optimal transformer performance?
Read along to learn more!
What is Ronin AI and how does it help the transformer industry?
Seetalabs’ Ronin is an AI powered tool to predict the THI score of each transformer using condition monitoring data for predictive maintenance. Ronin AI is uses machine learning (ML) algorithms to predict individual THI score for each asset or for a standalone asset over time, as necessary. In fact, the use of THI algorithms and its efficacy in improving asset management and grid performance is a hot-topic for the power industry. Other discussions are also available to showcase the benefits of using THI for original equipment manufacturers (OEM) and particularly, for non-expert industrial asset managers such as in the steel industry.
What are the benefits of using Ronin AI?
Ronin’s biggest value proposition is its triangular balance between time, cost, and accuracy of results. It is 10X cheaper than existing software suites with prognostics. It offers THI prediction in less than a minute and can analyze up to 500 transformers in a single shot. The training database of Ronin AI is non-conflicting, huge, and heterogenous. This means no overfitting and more accuracy. And we have saved the best for the last. It barely requires 30 minutes to train the asset manager on the use of Ronin AI dashboard.
How to use Ronin AI for maximum efficiency?
If the end goal is to predict availability, then assets with THI scores below 30% (very poor) category are the most alarming. Poor THI score could be due to high concentrations of dissolved gases in oil (obvious fault indicators) or poor oil quality. If THI score is low due to high level of dissolved gas concentration, then the asset manager can increase the frequency of DGA testing. In cases where transformers are equipped with online DGA monitors, it becomes rather easy and almost immediate to rank transformers and plan intervention. Simply put, if the asset manager’s end goal is clear then the efficacy of THI algorithm and effectiveness of Ronin AI improves multiple folds.
Best practices for using Ronin AI for transformer fleet management
A tool is simply math and algorithms. However, it becomes a pedestal to ensure optimal transformer performance if one understands the math behind it and the logic of algorithms. Ronin AI may look like a simple dashboard on the outside. But within it a huge, interactive, AI-powered framework that allows one to reduce the complexity behind data. For someone who is not a condition monitoring and diagnostics expert, the efficacy of any tool is ultimately to help make decisions. These decisions are usually based on urgency, which in turn, are ruled by logic. Furthermore, in a world full of ageing transformer fleets, it is often difficult to obtain the gold standard of data. Thus, an additional expectation from the software service provider is handle missing, inconsistent or sporadic data and turning that in to insightful results.
Ronin AI works best with CSV files containing at least the dissolved gases prescribed. The accuracy and precision moves multifold if all data is available. However, if some data is missing, Ronin AI uses the best recognition models to curate the most potential health index scenario and predicts the score. The confidence of such predictions are within a limit of 60-70% and can improve greatly if larger datasets, such as for fleets are uploaded, instead of a standalone transformer.
So don’t wait and grab your free trial right away!