Transformer Thermal Model - Our Vision

[thijsvanderhoeven301]

This post defines the vision for TTM: why it exists, who it is for, and the principles that guide how we build and evolve it. As team DALi, we hope this gives direction for everyone who uses or contributes to the model.

---

We believe in making optimal use of our grid, so that electricity can remain as affordable, reliable, and accessible as possible.

We will achieve this by:

• working in a modular way and breaking down major challenges into smaller solutions;
• implementing these solutions in a pragmatic and effective manner;
• contributing to the assessment of the trade-off between acceptable risk and customer satisfaction;
• initiating and driving national collaboration in this field;
• continuing to approach problem statements with our technical expertise and experience;
• Determining the thermal capacity of our assets.

What does our transformer thermal model deliver that aligns with this way of working?

Modular working

• Single-responsibility principle: facilitating risk assessment when thermally loading transformers beyond nominal conditions;
• extensive interfacing for specific transformer configurations;
• simulation of top-oil and hot-spot temperatures in transformers;
• a standard interface with (IEC-standard-based) standard transformer configurations for conservative modelling in the absence of asset data.

Contribution to risk management and customer satisfaction

• determining paper ageing within a transformer.

National collaboration

• by making the package open-source;
• by being active on our open-source repository;
• clear documentation with simple, step-by-step code examples;
• a toolbox that facilitates integration between commonly used tools (e.g. pandas) and our package.

Technical expertise

• measurements and reporting on assets in the field;
• validation of the model through load tests;
• IEC-standard-based hot-spot factor calibration as a fallback method when measurements are unavailable.

Pragmatic and effective

• vectorisation and short computation times;
• minimal use of large-scale AI in and around the product;
• continuous quality checks on the code;
• a methodology based on an internationally recognised standard (IEC 60076-7);
• input based on FAT measurement data;
• “good” is good enough.