Thanks for the nice article and drawing my attention to the IEA publication. I also look forward to reading about your models in more detail.

Agent based simulation is quite appealing, with the only caveat that, given the dynamics of all agents involved, simulation artefacts can arise which would not be seen on a state models with aggregate variables.

It is good to see the inclusion of feedback loops in your models, even though whether they are vicious or virtuous will depend on the loop gain and phase, and of course if there are any non-linearities creating limit cycles.