Also today we have posted a much longer article … ‘An interim ‘Conference Wrap’ (Part 2) from 2026 winter Conference Season’
In the compilation of that article, I added the question ‘How many years of reference data are required to construct a robust model of the future?’ … but thought it was worth separately collating these thoughts, in part because we’re likely to refer back to them later.
Here’s Paul Simshauser speaking at the AIE Event in May 2026, highlighting that his modelling utilises 80 years of weather profiles to produce 80 years of modelled renewable energy output:
… and here’s Oliver Nunn speaking at the Q3 seminar in Brisbane, highlighting that he also understands the need to model 45 discrete weather years (with an identical plant mix in all cases), to reveal the diversity of possible outcomes.
As we can see in the distribution plot below:
1) there are clearly some diversity between outcomes, depending on weather patterns
2) with some of the worst weather years being more recent years in the 45-year history … which feeds into questions about how climate change has already impacted weather patterns and is expected to do so more in future.
Also work reminding readers here that we included a 28-page Appendix 27, focused on ‘Exploring Wind Diversity’ in the GenInsights21 release, and shared this ‘precis of the ‘Exploring Wind Diversity’ analysis completed in Appendix 27 of GenInsights21’:
In all three cases, readers can see that a large number of reference years were used (in simple terms, as many as possible).
Looking at the slide showing unserved energy Vs reference year, it seems to me that the ESOO model would get much higher average USE if it uses 23 reference years (ie, what it currently does from FY2003-FY2025), than if it used 45 reference years. The current ESOO uses all 4 of the highest USE years, and 7 of the top 10. Using the additional 22 reference years will just drag down the average, making it much less conservative.