Artificial intelligence and waste avoidance

Effective energy waste avoidance relies crucially on the comparison of actual and ‘expected’ consumptions. Classically we do this on a weekly or monthly basis, using models for expected consumption that are linked to independent driving factors. But there are other ways to skin that cat.

Buildings will in many cases have a characteristic diurnal pattern of demand that can be expressed as a profile at, say, half-hourly intervals. With a large enough group of similar buildings, and taking account of drivers like the weather, it seems possible in theory to create a dynamic template for each building against which its demand can be assessed in near-real-time. The template is just a different way of calculating and expressing expected consumption, but it creates the realistic prospect of daily exception reports. Of course the implied excess costs need to be taken into account, because you need to be able to suppress the clutter of insignificant deviations, prioritise cases for investigation and estimate the value of resolving them, just as you would if you were using a weekly or monthly overspend league table.

The role of artificial intelligence here is to learn what ‘correct’ behaviour looks like and one advantage of this in large estates is that it obviates the need for human analysts to calibrate degree-day regression models for every meter. Another benefit would be the recognition of common abnormalities in profiles. Properly trained with correct human feedback, an AI-based pattern recognition system could in principle recognise symptoms that have occurred before elsewhere and associate them with remedies that have previously been successfully applied.

A further benefit is advanced benchmarking. In classical M&T we know that buildings can be benchmarked by comparing the slopes and intercepts respectively of their degree-day regression lines. A pattern-analysis system can take this more incisive analysis to a whole new level.

I will be interviewing James Ferguson, a keen proponent of AI in energy waste detection, on 15 July 2021 in my “Energy Conversations” series of open video calls. If this is a subject which interests you you canĀ  request a place in the audience here.