Conserve It’s Head of R&D, Michael Berger, recently presented "A Real-World Application of Machine Learning and Optimal Control Simulation for Chilled Water Plants" at the Australasian Building Simulation Conference.
In commercial buildings, the chilled water plant is typically the most energy-intensive centralised component, therefore decreasing its energy consumption is key to improving the energy efficiency of buildings and contributing to reducing CO2 emissions. The complexity of these systems has grown in the past few decades, and today numerous variables that have an impact on the overall plant energy usage can be adjusted during operation.
In his presentation, Berger proposes an optimal control strategy to select these variables with the objective to minimise energy consumption. The proposed approach relies on developing accurate machine learning models of key equipment within the plant. These models are used to run simulations in real-time to predict the energy consumption of the plant and find the control commands that minimise it holistically.
The solution has been deployed in real buildings on embedded computers and has achieved demonstrable energy savings on-site without relying on reducing the cooling production of the plant, therefore without compromising on space comfort.