Added Advanced Prices to intervals on /prices and /prices/current endpoints #214
Replies: 7 comments 15 replies
-
|
Thank You. Has any analysis been done on where the typical actual price sits compared to the advanced price range ? e.g. is it typically nearer the low, or typically the average of the two or a random distribution ? I'm guessing the answer should vary by state too. |
Beta Was this translation helpful? Give feedback.
-
|
I'll ask the data team, but generally our forecasts are more conservative, so towards the lower side. |
Beta Was this translation helpful? Give feedback.
-
|
Great news, we have been waiting for this enhancement! I will be injecting these values into EMHASS tonight. |
Beta Was this translation helpful? Give feedback.
-
|
what value will you choose relative to the low & high ?, or is it able to learn from both ? I'm toying with taking the difference between the two (the range) and then use the low + 10% x range. That'll give me something near the low. |
Beta Was this translation helpful? Give feedback.
-
|
Nice, these are coming though on both the 5 minute and 30 minute forecasts: |
Beta Was this translation helpful? Give feedback.
-
|
Have this injecting the low Advanced Price Forecast into EMHASS, let see how this performs. Purple line below (of course) |
Beta Was this translation helpful? Give feedback.
-
|
This is great news! Is there any updates required for the home assistant integration to use these advanced prices values? |
Beta Was this translation helpful? Give feedback.
-
|
Yep - I need to update the code. |
Beta Was this translation helpful? Give feedback.
-
|
:) you mean that stuff doesn't just happen? |
Beta Was this translation helpful? Give feedback.
-
|
Looking like the "low" number is pretty good for SA during a negative period.
|
Beta Was this translation helpful? Give feedback.
-
|
I've been playing with different formulae for choosing a price given we now have 3: low, high, aemo. I only have just over a day of data to work with, so things will change over time. Right now, the best results (on SA prices) has been to choose: a) aemo if it sits between low and high. Comparing low, high and aemo outcomes, they've roughly shared the same probability of being the nearest to the final price, so each around 33% "best" rate. However if you look at how often they are the worst choice, aemo is worst around 54% of the time, "low" is wrong 27% of the time, "high" 19%. Since Amber favours "low" I've compared the outcome of replacing "low" with the formula above and it currently wins 50% of the time and is worst 27%, so a big improvement in the wins column. None of the options predicted SA's mini spike this afternoon when solar dropped suddenly at the same time as wind was reducing. We'd need really good weather predictions to catch those ahead of time. |
Beta Was this translation helpful? Give feedback.
-
|
After 50 hours of data: Wins: low 14%, high 41%, aemo 45%. My formula is still winning out as a replacement for low, giving: 39% wins, 19% worst. |
Beta Was this translation helpful? Give feedback.
-
|
Definitely seeing an advantage to choosing the aemo price if it sits between low and high. Getting a 43% win versus 21% worst case, with a simple formula of, pick the middle value. This can be interpreted as the aemo price being good unless it's outside the Amber forecast band. |
Beta Was this translation helpful? Give feedback.
-
|
Also has an interesting treatment of CurrentInterval, Amber continue to publish low/high forecasts for the first 25 minutes of each interval and then remove those forecasts for the last 5 minutes as it is no longer an estimate and the full 30 minute price is locked in. |
Beta Was this translation helpful? Give feedback.
-
|
Let me know if that is annoying. I’m doing the HA integration mods at the moment and I feel like it could be annoying. Though I’m not going to add them to 5 minute current prices. |
Beta Was this translation helpful? Give feedback.
-
|
It does complicate the automation having it disappear at minute 25, when accessing via the API. It probably doesn't matter though for the integration, having the currentInterval in one sensor works well with or without AdvancedPrice attributes. The forecast sensors should include all AdvancedPrice attributes. I note the API can return the full data set from AEMO, upto 72 intervals (0330 for the next day), but the AdvancedPrice attributes are only available for the next 48 or even 49 intervals (24 hours). Is there a possibility to extend the AdvancedPrice forecasts to match the full AEMO intervals? |
Beta Was this translation helpful? Give feedback.
-
|
I'm a bit confused about Advanced Prices. What is the difference between 'predicted' and 'perKwh'? { |
Beta Was this translation helpful? Give feedback.
-
|
|
Beta Was this translation helpful? Give feedback.
-
|
My assumption is that Amber uses AEMO forecasts to help predict the price. Is the only difference that Amber combines a few models to improve the prediction? Or that the first AEMO current market forecast doesn't include some variables (such as network tariffs, etc) Just a bit confused because it seems like the app is using the perKwh value for the forecast. Wouldn't you guys use the predicted price if it's better? |
Beta Was this translation helpful? Give feedback.
-
Correct.
AEMO's “forecast” isn't really a forecast — it's the current state of the system. ie, if the forecasted operational demand doesn't change, then the price won't change. There are models that go into each generator's models, but the price is a second order effect.
It's complicated. We use it for Smartshift, but we made the decision at this point to show the AEMO prices for the moment. We'll be updating the UI at some point to show our forecasts too, but we aren't really sure what that'll look like yet. |
Beta Was this translation helpful? Give feedback.
-
|
Awesome, thanks for the detailed reply. So essentially if I am looking for a prediction, in your opinion, the predicted price is likely to be a better choice? (I know no guarantees) |
Beta Was this translation helpful? Give feedback.
-
|
Yeah, I'd rely more on the predicted attribute. It's currently a little too conservative, but we are continually tuning it. |
Beta Was this translation helpful? Give feedback.
-
Hi everyone, my name is Marty and I have recently joined Amber as a software engineer working alongside @madpilot and team.
We have rolled out Advanced Prices!
What are Advanced Prices?
The
AdvancedPricerepresents the Amber forecasted price prediction for a given site. These values have been "priced" for the givensiteId, meaning that it takes the Amber forecasted wholesale price, and includes network and market fees.Where is this information included?
This change will add the
advancedPriceattribute into theCURRENTandFORECASTintervals returned from the amber public API endpoints:/sites/{siteId}/prices/sites/{siteId}/prices/current.Also note that it is not returned for the following (given these are not estimates).
ACTUALintervalsCURRENTintervals whereestimate=falseAdditional Information
The OpenAPI docs have been updated, and are available here.
The definition of the AdvancedPrice property is as below:
An example output of the intervals returned from
prices/currentis below:Beta Was this translation helpful? Give feedback.
All reactions