Reactive Technologies has launched ‘CHP optimisation as a service’ which it says uses machine learning to optimise run times and revenues for combined heat and power plant.
The firm’s initial client is a large retailer, for whom it is optimising “hundreds of small CHPs” around the country, according to head of retail, Mark Cavill, who recently joined the company from Engie, where he was head of demand-side response.
Cavill says the service takes in operational data, hashes it with industry data via Reactive Technologies’ “machine learning algorithm” and works out the best way to run the CHPs for both optimal efficiency and revenue generation.
“A significant number of smaller CHPs were installed in the late nineties and early noughties when the technology basically assisted with planning applications,” he told The Energyst.
“Back then, gas prices were quite low and the spread between gas and power prices enabled CHP to really stack up.”
Today that price spread is narrower, eroding the marginal benefit of running CHP.
Equally, building operations have changed over the last decade or two, while CHP operational strategies may not have kept pace, or been changed at all, says Cavill.
“As a result, many CHPs may be running when it is not economic to do so, or vice versa.”
Data-driven heat and power
The service takes in weather data, BMS data and forecasted heat demand as well as power prices and third party charges.
“Based on all this data [for the retailers’ CHPs], we created an automated process to capture all of the electricity price information specific to its contracts and sites, plus forecasted heat demand on each site, while looking closely at efficiency and operation of each CHP asset,” Cavill explains. “Then we created an algorithm that we run to produce a schedule of when it is most economic to run each CHP.”
That schedule includes running the CHPs within various flexibility programmes or to facilitate peak network charge avoidance.
“The resultant savings range from 10% to more than 30%,” claims Cavill.
While the service currently produces the schedule, Cavill says “phase two is to install controls to remotely control assets”, so maintenance teams do not have to input recommended run patterns.
Cavill says the service can work with “single massive CHPs down to hundreds of smaller distributed assets”.
This article originally appeared in The Energyst December/January print issue. If you are involved in energy procurement, management or flexibility within your organisation, you probably qualify for a free subscription.