Machine-learning model pinpoints dying power grid components

Machine learning could one day help energy providers better pinpoint failing or compromised components in power grids, or better identify traffic congestion for local authorities, according to a study. A research project led by MIT describes a technique capable of modelling complex interconnected systems made up of numerous …

[b0llchit]

Promises

Machine learning could one day help...

This opening sentence is typical for the whole AI/ML saga. Every year we are promised that soon it will be ready to do fantastic things. Soon it will be better and no longer make mistakes. Well, maybe, just maybe it could one day be sentient. It could one day decide to kill us all and be done with it.

Most of the AI/ML stuff feels like a sales pitch. Please invest your $€£ to make the better world of tomorrow. Your reward is awaiting. Soon we will be ready. Today we will take your money.

[Korev]

Re: Promises

So what you're saying is that at the current time it's not much use

[b0llchit]

Re: Promises

No, I'm waiting for it to become sentient and kill us all. Then we no longer need these discussions and the planet can move on to the next stage of evolution.

[Cederic]

Re: Promises

Meanwhile people are using Alexa and Google's thingy and Siri on a daily basis.

Companies are successfully using ML based algorithms to drive profit.

Smartphones create fantastic images through computational enhancement.

It's almost as though society is already benefiting from AI/ML.

[John Brown (no body)]

Re: Promises

Except of course, when it goes wrong. Then it's "no ones" fault because "the computer made a mistake". And it's cheaper. Especially when it involves monitoring a large and diverse network such as power distribution. The supply companies will be all over this if it helps their bottom line, even if it's only "maybe, nearly,,not quite as good" as the current systems.

[Old Used Programmer]

Is that you HAL?

So long as it doesn't make you go out into space to replace a "failing" component that is not--in fact--failing...

[Gene Cash]

AI is more bullshit than fusion

At least fusion has real science behind it.

[Anonymous Coward]

As one close to the power quality measurement and modelling done in the UK, I can say with some certainty that this is an impressive capability if it remotely does what it says on the tin. Hunting for signs of harmonic formation is non trivial on a simple system let alone on a complex meshed network like China or the UK.

The main problem is deploying and maintaining instrumentation without shutting down the grid - for the cost of minor outages for short duration utterly dwarfs the ability to accept such outages and costs linked to them.

Availability and reliability are strange metrics; achieving maximum availability (zero maintenance) runs counter to maximum reliability (100% maintenance).

[Kevin McMurtrie]

Training

I imagine PG&E using this and having it flag repairs as extreme anomalies. Sure, they cover some rugged terrain, but they also have disasters where extreme weather means adding a hoodie to your shirt and jeans before going for a walk.