Time 

"I see a girl, wondering why, her heartbeat can't keep up,

I see a boy, wondering why, nobody figured it out."  

- Princess Chelsea, Time

That's it, I'm doing it. I've long since wanted to name all of my blog posts after Princess Chelsea song titles and/or lyrics.

And so that's what I will try to do, for the foreseeable future.

Challenges of AI in Renewable Energy

Now I'm moving on to the second section of the EY article on AI and Renewable Energy. This section is called: "The challenges of applying AI across the sector".

Although the impact of AI across the renewable energy sector is undoubtedly going to be huge, there are many many challenges associated with its integration.

The Vulnerabilities and Risks of AI in the Renewable Energy Sector

It is concerning that relying on AI too much in the renewable energy sector could leave us vulnerable to cyber attacks.

As someone who has been researching these things for months - I can definitely confirm that cybersecurity is one of the main concerns of the smart grid in general. It is incredibly important to protect smart infrastructure from cybersecurity attacks.

Cyber security attacks in the smart grid can be incredibly carefully coordinated, and can take months to prepare.

Examples of Cybersecurity Attacks in Renewable Energy & AI

Historically, there have been examples of cyber security attacks involving "exploiting vulnerabilities in firewall firmware". Whatever that means, sorry.

However it seems that the technology is becoming increasingly progressive such that cyber attacks are harder and harder to achieve.

Hacking into an OT (operational technology) system which is typically what is used for these kinds of things would require actually knowing how to work it, which apparently is quite hard. "if hackers did get into operations networks, they would need to learn the equipment and setups." So therefore, as manipulating equipment and setups can require real expertise, these networks are apparently not at such great risk anymore.

Performance Issues

If bad data comes into the AI algorithm, then the outputs will be bad as well. The article says here that: "it is critical that data is taken and made machine readable, so that it is quality in, quality out." Frequent verification of data in order for AI to be really trusted. This is the hard part about AI, I suppose.

There is also the bit about data bias, and avoiding it, and the massive risk of that. There may also be risks where AI has never encountered an issue before and therefore it has no more historical data on that. All of this, however, can be overcome (according to the article):

• By using different data sources
• By using more data sources
• By being more selective about your data sources
• By using different techniques.

Not sure what the last one means, sorry. 

Not sure what kinds of different techniques can be used...

A Lack Of Reliable Connectivity

One of the major barriers to the uptake of this kind of technology is a lack of the right kind of network coverage. This article refers to it as "cellular network." At work, when I worked for a major UK renewable energy supplier, we referred to it as HAN or WAN - Home Area Network, and Wide-Area Network.