I can speak for the GB case. Low Frequency Demand Disconnection (LFDD) occurs automatically and in stages when the frequency drops until it stabilises. The substations or feeders that are tripped off are not currently determined by real-time metering - instead they are pre-allocated based on their typical demand. This means that the system operator does not really know how much demand will be disconnected at any given time. If it's sunny, you could easily trip off a lot of solar generation connected on the low voltage network, causing the frequency to drop further. It is far from optimal!
> The substations or feeders that are tripped off are not currently determined by real-time metering - instead they are pre-allocated based on their typical demand. This means that the system operator does not really know how much demand will be disconnected at any given time.
This is wild. From a amateur technical perspective, it would only take a cheap hall sensor inside the transformer to have a pretty good guess of how much current has been flowing to the load.
Hell, put the hall sensor onto a board with a micro controller and a LORA transmitter and stick it to the outside of the feed line. Seems like an incredibly cheap upgrade to get real-time load data from every substation.
The nice thing about frequency based regulation is it's an inherent property of the system, so as long as you're connected to the grid you've got the info you need to decide when to turn on or off.
If you're monitoring real time power consumption you then need a whole extra infrastructure to communicate this info back and forth. Of course you then have to consider how you're going to keep that extra infra online in the event of power issues.
Frequency based regulation is only telling you that something is wrong, but not what or how to fix it.
If you find yourself in the middle of a black swan event, and 15 GW have tripped offline, you have milliseconds to dump pretty much exactly 15 GW of load, otherwise more generating capacity is going to trip offline very quickly.
If you only dump 14 GW because you used historical data (which happens to be imprecise, because today's cloud cover reduced rooftop solar output), you're still going to be in trouble. A detector scheme with sensors at every substation would allow you to do just that.
The board is not the expensive part. It's the getting reliability qualified and then having staff fit it to every substation, arrange the data links, and construct the dashboard.
I also wonder what the realtime requirement is. Data from a minute ago is fine .. except in this kind of situation, when things are changing very quickly.
Doing that at scale is tricky and requires a lot of people to participate in the mechanism, whereas the law only forces producers above a given size to participate.
The estimates we get from seasonal studies are usually close enough, especially since load shedding isn't a finesse exercise.
The situations that require load shedding usually give operators only a few minutes to react, where analyzing the issue and determining a course of action takes the lion's share. Once you're there, you want the actual action to be as simple as possible, not factor in many details.
What's the comparison with conventional attention using a more aggressive (lower temperature) softmax? I can imagine that for the multi-needle retrieval test this may also give a performance boost, although at some cost other more creative tasks.
I had the same thought: Just eye-balling the graphs, the result of the subtraction looks very close to just reducing the temperature.
They're effectively doing softmax with a fixed temperature, but it's unclear that this work is going to do better than just learning a per-head temperature parameter.
The other way to think about this is that it looks like a hacked-up kinda-sorta gated attention. If that's the case, then doing softmax(alphaq_1k_1^T - log_sigmoid(betaq_2k_2^T)) might be better? (where alpha,beta are learned temperatures).
The vast majority of wind that is unharvested is curtailed due to constraints on the transmission network between Scotland (lots of wind) and England (lots of demand). There is a significant amount of pumped hydro (and battery storage) in Scotland to help in these instances but there are still enormous costs associated with curtailment (approx. £1 billion per year).
Often the high costs of "curtailment" are inflated by counting the cost of gas plants in England that run to fill in for transmission blockages.
Of course, if the previous government hadn't effectively banned onshore wind in England, the cheapest source of energy available, you'd be able to deliver that energy for a quarter of the cost.
And, for extra added irony, this is an occasion that the classic "the wind doesn't always blow" line backfires, as either a) if it's windy in Scotland then it will be windy in England avoiding the gas cost or, b) the wind across that distance is decorrelated and you can displace even more expensive gas from the grid.
Pumped Hydro and battery storage are a drop in the ocean of the amount of storage needed to secure the grid, especially as people shift to electric cars.
I take the opposite view - as more people have electric cars there are more batteries available. With some sort of dynamic pricing, and bearing in mind most commuter mileage needs only 1 full charge a week, you could encourage consumers to charge up when it's windy and/or hold off charging when it's not, so better matching demand to supply and reducing curtailment.
This is already a thing in some places. In Norway we have spot pricing of electricity, market price hour by hour. This means that it's possible to have a contract that lets you pay the market price at the time you use the electricity (timespotavtale). Even with the usual contract (spotprisavtale) where you pay the average price each day you can see that price (set by Nordpool) by looking on line or by subscribing to a service that notifies you when the price drops and decide whether or not to charge the car.
The Nordpool price today in the region where I live is 0.80 NOK/kWh, that's 0.074 GBP/kWh or 0.056 USD/kWh.
How many people actually do this and are happy with it though?
Here in the Netherlands there are also contracts like that, but people that are interested in them tend to put a lot of time and effort in then tracking those prices. I don't see regular consumers as a whole ever being interested in that. For now, the benefits are also small, and I happily pay a few percent extra (net) to not have to bother.
I also feel like it's not a good direction to move it: consumer energy markets are actually quite predictable so taking out long term contracts should stay the norm (consumer contracts and supplier contracts). Using car batteries as storage for solar can be addressed better by making it more attractive to charge at work, where I assume most park their EVs in the day, but right now pay full price despite providing a place to store surplus.
Especially, because most energy used is currently not electricity:it fossil fuels. We need 4x for a zero-carbon grid and to be secure we need over a week's stored electricity: 10s of TWh. The plan for the UK grid has 50 Two in it. Will people really postpone charging their car for a week if they hit a cold still week in January?
If everyone shifts to EVs and Vehicle to Grid becomes a normal thing then securing the grid would be a solved problem. Most cars spend more than 90% of the time standing still. If every parking space was fitted with a mid-range, say 7 KW charging point that would go a long way to providing sufficient grid storage.
My Tesla S 70D could supply my house for at least 24 hours if it had V2G.
Here in Norway we have 20% of private cars electrified. No impact on the grid and not a lot on total electricity use. If all energy used in the transport sector were to switch to electricity today it would add about 50% to the demand for electricity. But EVs are about 90% efficient versus about 25% or less for ICE vehicles so in fact it would add less than 15% to the electricity demand.
My first thought on seeing the PCA embeddings scatterplot was "I wonder what pdfs are at the centre of those two clusters?" The most typical pdfs on the internet.
