We are at a crossroads in the energy and infrastructure business, wondering whether the electricity networks will become even more connected at the transmission level in supergrids spanning the length and breadth of Europe and beyond, or whether energy production and consumption return to being more local (which is how it all began...). Both outcomes are likely, not one exclusively. Optimal solutions are rarely one extreme, but how do we soften our tendency for tribalism, for defending our pet solution? How do we realize our role in the whole?
Judicious use of computers will help.
My colleague Eero Saarijärvi has written in his forthcoming doctoral thesis, algorithms are “unbribable”, but he also points out the care need in parameterization, or the computer will get things wildly wrong...
We have taken the issue of parameterization seriously in a network planning algorithm we have been working on at Aalto University, but that leads to a potential risk. If the user myopically specifies load growth, customer interruption costs and a whole host of other parameters specific to each secondary substation or even LV customer up to a distant planning horizon, the resulting network plans may not be appropriate if the forecasting is inaccurate. Added to this are the challenges brought about by the hard-to-predict production and viability of renewable distributed generation, which we are presently modeling from a planning perspective.
It is hard keeping up with developments, and equally challenging taking this research area to a new level, a bit like sailing upstream with the wind against you. Nevertheless, we have tried to take this decades old research topic, which has benefited from significant contributions from Finland (three professors come to mind: Jarmo Partanen, Erkki Lakervi and Matti Lehtonen) to what qualifies as a ‘new’ set of incremental contributions.
The best compromise we have come up with so far, is to use the algorithm in a two-stage format, taking full account of existing network, every year or so planning towards an intermediate time horizon where forecasting should be reasonably accurate and a ‘moving-target’ on the distant planning horizon. This means annual use of the algorithm, to ascertain a close-to-optimum plan for building new network and upgrading old, with the user exercising judgement as to the on-the-ground details and the shifting of close-to-optimum theoretical guidelines to fit annual budget constraints.
In that sense, we are happy to have been involved in developing an algorithm that will not steal jobs, but will hopefully release the energy and time of experienced distribution network planners to carefully parameterize and judiciously consider and fine-tune the results, letting computers do what they do best.