> Autobidder is a real-time

it observes data and makes decisions continuously

> trading and control platform

it makes trades on the power markets as well as controls your battery

> that provides value-based

you can control a model of fundamental economic value embedded in the software, so your decisions can be made not just based on observed market prices, but also your own opinion of what things ought to be worth

> asset management and portfolio optimization,

you view batteries as financial assets and look for ways to optimally manage them to extract the most money, and you can view and manage multiple batteries as a single economic unit

> enabling owners and operators to configure operational strategies that maximize revenue based on their business objectives and risk preferences.

you can choose to have more money with more risk, or less money with less risk

>> Autobidder is a real-time

>it observes data and makes decisions continuously

The correct term for this is "online", not "real-time" which has another meaning.

My understanding of the terms is that online means it processes streams of data, but not that it is fast enough to keep up with the real world. Realtime means that it is fast enough to keep up with the real world. Realtime seems to be the appropriate term to me.

Do you have a different set of definitions?

Yes this is the correct term. Realtime processing in the operating system context vs the streaming data context are different things.

https://blog.syncsort.com/2015/11/big-data/the-difference-be...

Realtime software often means "hard real time"[1] which is a constraint where a response is guaranteed within a particular time.

This is a fairly specific area where special real-time deterministic operating systems are used etc.

I think your usage was fine, and the pedantic use of "real time" is dying anyway. There's a whole field of "soft real time" which blends into online, and some parts of the battery control SCADA systems may well be true real time anyway.

[1] https://en.wikipedia.org/wiki/Real-time_computing

I was just responding to the definition given. "observes data and makes decisions continuously" is not a proper expansion of "real-time".

The time scales it operates in are usually quite a bit larger than systems that would typically be discussed here, but I don't see how "online" would be a descriptor than "real-time".

It's software that decides when to charge the battery / buy power and discharge the battery / sell power using forecasts based on historical demand trends and the spot price of power.

Technical innards of the Autobidder system in case you care to know more https://youtu.be/ggdYts4muu0

It’s basically a market forecasting system which allows the operator to place bids on various energy markets such as supply, contingency and ancillary services. The battery has limited resources (power output and energy storage) so it needs to plan well ahead for making the most money for each watt-hour it buys or sells.

Autobidder uses ML and other signals to maximize the revenue for asset owners based on electric market data (real time and historical). It has a rough idea when the grid will be demanding power, how fast the storage project can charge and discharge, what the minimum aggregate state of charge should be, etc.

Not affiliated with Tesla except as an investor, have gotten to test drive Autobidder as part of an engineering/procurement/construction storage project.

What are the industry incentives for Tesla to do this, instead of utilities?

It just seems kind of weird for Tesla to be saying "We'll sell you this thing you tell me you need" and also "Here's us building a thing to help you tell what you need."

> What are the industry incentives for Tesla to do this, instead of utilities?

1. Most utilities operate power plants, they don't operate power storage facilities. From a capital efficiency standpoint, power plants want to be run all the time, so there's not much that they can do that's particularly fancy.

In contrast, Tesla both buys and sells electricity, so they have the opportunity to arbitrage across time. This is especially true in places where the cost of electricity dips below 0.

2. There are some utilities that have stored power (e.g. hydro dams). I suspect that they don't employ the engineering talent to pull off something like Autobidder.

But those energy providers ("utilities") have the necessary political and budget power to pull off something like Autobidder. In many countries they are controlling local governments. Buying such a logistical solution is peanuts for them, automation is everything.

They are already doing better things than just dumb ML, they have proper models and constraint and planning software.

If they had the will and resources to do it, they would’ve done it. Having been in both engineering first and engineering last orgs, it’s clear they cannot execute in this regard (like a tech company).

Right. Because these utilities have, as Ruban says, "political and budget power" they are fat and happy, and don't bother with engineering innovation.

To be fair, in many cases markets are explicitly set up to preclude innovation.

Monopoly grants with rate caps don't afford much aspirational budget for SV-wage engineers.

Do you know any open source alternatives? I've just signed up for power using a plan to pay using spot price. The provider already does show forecast for the day, but I'm sure they don't have an API.

Edit: Well, seems someone mitm'd their app to find API. Can plug that into something that can analyse entire NZ's trends.