# How about that weather year eh?

> Most models run one weather year on repeat. We run up to fifteen, because there is no typical year, and the more renewables we build the more the weather determines the price.

# How about that weather year eh?

Most models run one weather year on repeat. You may not even know which. We run up to fifteen, because there is no typical year, and the more renewables we build the more the weather determines the price.

> **INFO:** TLDR
>
> * A *weather year* is a real past year of demand, wind and solar shapes based on real weather conditions, mapped onto the future so the model knows what the weather does at each half-hour.
> * Pick one and repeat it and you get groundhog day: the same summer, the same winter, the same wind speeds, the same clouds, forever.
> * Do you know which weather year you use? Going all in one weather year is a risky strategy.
> * Picked a good one for your specific asset? You may not see those returns in reality. Picked a bad one? Your asset may not get off the ground.
> * We're happy to help you on your journey to improve your valuation processes.

## What is a weather year?

A weather year, or reference year, is a projected set of future conditions based upon a historic year's weather conditions. Because we don't know what the weather will be next year, or in 2050, we take historic weather patterns (or weather years) and map those conditions forward. Those conditions impact demand shape and solar and wind output. For a full explanation of weather years, see the 'Weather years 101' section below. Before, we're going to delve into why weather years matter so much.

## The typical approaches

Price forecasting and valuations quality typically falls into four buckets:

| Quality | Approach | What it means for your valuation |
| --- | --- | --- |
| ●○○○ | **No weather correlation at all** | You buy a price forecast from a consultant, then multiply it by a generation trace from somewhere else. Unless the forecast came with its weather assumptions, those two are not describing the same day. Still fairly common - and you'll likely be significantly overvaluing your VRE. |
| ●●○○ | **One repeated weather year** | You know your weather correlation, but you run the same year over and over - the groundhog-day problem, see the next section. Also very common, and risky: it can unfairly make or break a project. |
| ●●●○ | **A range of weather years** | You model the range and know your earnings band. This is what we offer, and it puts you ahead of the rest. |
| ●●●● | **Monte Carlo across the weather years** | You value the asset over thousands of simulations that stay fundamentally correlated, so the wind, sun, demand and outages still belong to the same day. Statistics and fundamentals, combined sensibly. We also offer this. |

## The groundhog-day problem

The convention is to pick one weather year and run it for every year in the horizon. If you plot your price curves on a monthly basis, you may notice some repetitive behaviours. And there are winners and losers. In this example, assets that generate heavily in July benefit. Those that don't, lose out.

## How bad can it get?

I don't know what weather year I use or I only use one weather year? How bad is it?

Let's look at one wind farm in the New England renewable energy zone (N2, NSW), valued over five weather years on the exact same scenario.

Depending upon which weather year you used, the earnings can vary by 11% per annum (on the 5 weather year sample). Repeat that year after year and it compounds: from 2030 to 2040 the same farm earns **$258m** on the worst weather year against **$299m** on the best. A gap of **16%** (again, on 5 weather years). The more you run, the more of the range you actually see.

### Is it consistent across REZs? No!

Let's look at the results across every renewable energy zone (REZ) in the NEM. The earnings potential is the spread across weather years; the black diamonds are one single year, 2020. In Queensland, 2020 sits at the top of almost every zone. In South Australia and Victoria, the same year sits at the bottom. There is no universally good or bad weather year, which is exactly why diversification is so important to a functioning highly-renewable energy system.

### Does it matter for firming? Yes!

Swapping the weather year around doesn't impact the underlying energy price as much as it impacts the volatility. At Varial, we split the power price into an energy component and a volatility component.

For volatility, it has historically been driven by coal plant tripping, network outages or constraints, extreme heat days, or some combination. Going forward, we think volatility will depend more upon lengthy VRE droughts, as the previously mentioned issues (outages, constraints, heatwaves) are met by more storages. What they can't cover is running out of charge during VRE droughts.

Model a weather year without VRE droughts and everything is rosy. But you'll undervalue your firming assets.

## Weather years 101

To forecast a price you first have to decide what the weather does across every half-hour of the next thirty odd years. So the industry borrows it from years we have already lived through.

We use AEMO's published weather traces with the Integrated System Plan. For each historical financial year 2011 to 2025, a half-hourly demand shape for every subregion and a capacity factor for every renewable zone, is mapped forward over the 30-year forecast horizon.

The easiest way to understand them is to look at one sample week.

The same week averages **$24/MWh** on 2021's weather and **$114/MWh** on 2025's.

Under the hood, there are several changing variables: demand, solar supply, wind supply. We also choose to model coal availability on historic availability factors to replicate realistic outages. These factors impact storage behaviours and interconnector flows.

Demand shape: how hot or cold is it?

Wind: in 2021 it ran at 14% capacity factor across the whole week, against nearly 50% in a good year.

Solar: how cloudy is it?

We add historic coal availability too, applied forward onto the remaining plant, because outages have always been a primary driver of volatility.

These availability factors flow on to bidding, storage dispatch, interconnector flows, and ultimately, power prices.

## What next?

No matter where you are in your price forecasting and valuations journey, it is worth finding ways to improve. One weather year is not enough information to make significant decisions upon, and running the range turns 'we assumed a typical year' hope into 'here is the spread, and here is where your valuation sits in it'.

> **TIP:** Three questions worth asking your forecaster
>
> * **Which weather year is this built on?** If nobody can tell you, you can't check anything else on this list.
> * **Did my generation trace come from that same weather year?** If it didn't, your revenue is very likely overstated.
> * **What does my asset earn on the other nine?** That range is your risk, whether or not anyone has shown it to you.

We will happily provide as many weather years as you'd like in our valuations - it's as simple as picking from the set and pressing go. And we'll do it for the same price as the single weather year forecasters.

Want to see your own asset across every weather year? [Get in touch](/contact/), or [follow us on LinkedIn](https://www.linkedin.com/company/varialenergy) - we post this sort of analysis regularly.