Guest Blog: Michael Prager – What does AI mean for Energy Managers?

Artificial Intelligence, AI is a new form of computing in which the system is pre-programmed with logic that allows it to mimic the way that humans learn and perform cognitive functions. We see it now as speech recognition, optical character recognition and increasingly pattern recognition and early forms of logical reasoning.

Michael Prager, Chairman, Optimal Monitoring

Artificial Intelligence, AI is a new form of computing in which the system is pre-programmed with logic that allows it to mimic the way that humans learn and perform cognitive functions. We see it now as speech recognition, optical character recognition and increasingly pattern recognition and early forms of logical reasoning. (see Duncan Everett’s, ‘What the hell is AI’ blog). So instead of being programmed with rules that will always be hard coded and apply to all situations, without reference to what they are or why they happened AI will compare data it is seeing to patterns it ‘knows’ and make a logical inference from these two streams.

In energy that means two things:

  1. The starting point is ‘no normal’. That means there is no set level of good or bad performance. The system, ours is called EMMA AI and is the most advanced instance of AI in energy management to the best of our knowledge, will look at what’s normal for each site and establish that as the baseline at the same time as it will compare this ‘normal’ with other baselines in similar environments e.g. retail, manufacturing, health care and make a judgement on whether this is on, below or above the broader sector normal. In this respect EMMA AI is acting just like a human would in evaluating what is normal and how this compares to a comparator group.
  2. The AI system, EMMA AI, will take the consumption data that monitoring systems have always measured and group it into patterns that show spikes and troughs in usage but instead of just throwing masses of data back to the user to interpret it will compare these data to similar patterns and then draw from its knowledge base what the causes were of these anomalies and suggest what these might be. So, as a simple example: Electricity usage doesn’t reduce when the building empties after the workday. Looking at the level of consumption EMMA AI will compare this to typical levels of usage for lighting and heating/cooling and make a determination which it’s likely to be. If the usage was too low to power an HVAC system she will assume that this is lighting left on and make the suggestions, ‘check that any timers on displays or elsewhere are still in sync’, put a ‘last person to leave switch off’ sticker on all circuits not needed when the building is empty, ‘ Crucially she will also ask for feedback in simple plain text English. Were any of my suggestions correct? If not what was the case of this unexpected demand? In this way EMMA learns what was right, if something happened that she didn’t have in her knowledge base and continually learns the link between cause and effect so that her recommendations become ever more accurate and ever more localised to each site monitored over time.

The system learns from each experience rather than having knowledge programmed in a hard coded way. That’s AI.

Let me try to answer the questions that might be flying through your mind rather than just waffling on about the merits, myths and bunk around AI.

Will AI be widely used in the Energy sector any time soon?

Yes. EMMA AI is live now, others will follow no doubt. AI isn’t a coming technology, it’s already in our lives now. Siri uses speech recognition to let you control your iPhone or iPad by voice rather than a keyboard. Amazon Echo does much the same but from a much more diverse knowledge base. Chatbots and robotics use AI rather than rules based computing to determine what the question means and where things are respectively. Energy, with vast reams of data which are slow and expensive for humans to interpret, is a very good area for AI to add value to the process.

Does that mean my job will disappear?

Unlikely. There is no case in recent history of new technology destroying more jobs than it creates but it will probably change your job. Filing Clerks have gone away and returned as admin assistants or office managers. Ledger clerks and bookkeepers have given way to cost & revenue accountants. Flight engineers on aircraft have gone away but the demand for pilots has never been higher (so much so that BA pilots feel confident enough to go on strike for more pay). AI systems will take away the grunt work and if you only do grunt work like collecting and sorting masses of data on excel spreadsheets expect that to disappear (that should make you quite happy no?). Think about what you will be able to do with the knowledge that an AI system provides rather than how you might compete with it. There will be opportunities aplenty for those who can make this small mindset adjustment.

Will this help me realise efficiency savings?

Yes. Firstly by having the system suggest what the fix is to the problem or opportunity you will be able to send instructions straight to site or outsource provider; get a plumber to check the valves for a leak in the men’s toilets in building C. Have a boiler maintenance guy inspect the burners as they have started to use much more gas etc. The need to spend hours on figuring out what the data is telling you rather than acting on it will dramatically speed up resolution time across the board and from being able to oversee, say five sites, you will be able to do that for fifty sites on an AI platform

We only do capital projects with a guaranteed return.

In the words of Mrs. Brown. ‘That’s nice.’ As you well know there are three levels to energy efficiency and AI will make all of them more efficient:

a) Pick the low hanging fruit. Reset heating controls that have gone out of sync, see leaks in gas and water and get them fixed etc

b) Change behaviours. Have people switch off non-essential equipment when not being used e.g. gas burners in a commercial kitchen that are left permanently on because someone might come in and ask for an omelette! Or lighting in an empty building, or heating the building three hours too early etc

c) Capital projects like replacing old halogen lighting with LED’s. Installing heat exchangers etc

All of these contribute towards saving energy, reducing C02 and making the business and the planet a better place. The problem has been that in many situations, retail is a prime example, the cost of fixing lots of small issues has been greater than the savings that could be generated hence ‘we only do capital projects….’. AI will eliminate or dramatically reduce that cost so e.g. saving £10 a week on energy in an estate of 1,500 sites – is a pittance per site but do the maths £10*52*1,500 = £780,000. Worth doing and with AI the modelling that a reliable ROI analysis requires can be done way more accurately and using the efficiencies above at a much better ROI than previously so as an energy manager you will be far better placed to handle behaviour change as well as capital and the blindingly obvious efficiency issues than ever before.

We have already implemented a behavioural program, so we are good thanks

Well good for you if you have, but as we know, it’s really hard to keep this going, so maintaining savings made is tough, let alone building on them. In general people support what they help to create. By making the findings of your AI system available and even usable by your internal constituencies they become part of the solution rather than remaining part of the problem. Nobody wants data, they want solutions. (you want data, we know that and you will get it, it’s just the site level personnel for whom data = more work. Solutions as instructions on what to do = job made easier) . AI systems will give you proof positive of how the changes you and your teams make are effecting kwH and C02 . That can only be helpful.

We’re a small business. We won’t be able to afford an AI system

You are wrong. This is one of those examples of where the deployment of a new technology is actually reducing the cost of it. In general, and this is a wide definition so may have some variability but an AI energy system will cost around 1-2% of existing energy use and should generate 10-15% sustainable annual savings or efficiencies. For large multi-site enterprise the investment requirement at the moment for EMMA AI is a one-time user license of £5,000 per annum and £20 per meter. That surprised you didn’t it.

If you would like to know more you can contact Mike by emailing [email protected] or contact Optimal Monitoring on 01494435106.

Optimal Monitoring will  also be running a demo of EMMA, the world’s first AI Energy Manager, at Energy Live EXPO 2019 on the 5th of November at the  QEII Centre in Westminster. You can put EMMA through its paces by letting EMMA study your building’s energy usage, suggest recommendations and afterwards you will receive one to one feed back on the day.

If you are interested, get in touch with [email protected]

This is a promoted article.

Latest Podcast