Logo - CCME
Banner Main – Digital Issue

Harnessing Artificial Intelligence for maximising chiller plant efficiency

An intelligent optimisation system using AI algorithm built into its core is just what the doctor ordered for its ability to facilitate continuous, dynamic optimisation and, thus, deliver greater plant efficiencies, writes Rehan Shahid

| | Oct 28, 2020 | 5:33 pm
Share this story

In this article, I will focus on the advantages of having a chiller plant optimisation system on top of Building Automation System (BAS) and the role of Artificial Intelligence (AI) in achieving maximum plant efficiency, followed by what needs to be considered when opting for such systems.

But, let’s first look at why there is so much emphasis on the efficiencies and optimisation of central chiller plants. Some main reasons are as follows:

  • Ever-increasing energy costs
  • Carbon footprint
  • Cuts in O&M budgets

Considering that a chiller plant is regarded as the “heart” of an HVAC system in a building, the need to optimise it becomes critical to achieving better building performance. In other words, optimising a plant represents one of the largest energy-saving opportunities.

The key components of a chiller plant are cooling towers, chillers, pumps and filtration systems. From a total plant energy consumption point of view, among them, chillers typically consume the largest share of energy, followed by pumps and, then, cooling towers.

Now, ideal plant efficiency or the potential for efficiency is determined by the following design and operating decisions:

  • Real-time measurement and reporting
  • Full plant optimisation on top of BAS; this takes multiple variables into consideration, communicating with BAS and also directly with the plant components for information
  • Relational control approach
  • High-efficiency equipment
  • All variable-speed plant equipment
  • Predictive maintenance

In order to have optimised “excellent” plant performance (see Figure 2 – excellent band between 0.5 and 0.7 kW/ton), a holistic approach needs to be adopted to maximise efficiency. Therefore, all components, such as cooling towers, chillers, pumps, air-handling units (AHUs), fan-coil units (FCUs) and filtration systems are required to be included in the optimisation exercise, and this is where AI comes in.

AI can perform an extensive range of functions not purely restricted to usual concepts of IT; computer algorithms are increasingly able to almost instantaneously access vast amounts of data, compare and organise information and perform automated procedural and analytical functions. AI predictive modelling can even foresee a breakdown occurring, allowing ample time to fix the problem before it results in downtime.

Let us now look at the plant optimisation solution. Primarily, it has three main components:

  • Automation of system
  • Real-time measurement and management
  • Optimisation system – an additional layer of chiller plant optimisation software, preferably using AI to allow real-time data analysis, learning and learning from experience

Key criteria in plant optimisation

Plant optimisation should be a scalable and adaptable controls solution that takes an all-inclusive view of the entire plant, resulting in optimised operation of the complete system and minimised energy use. It is equally important to ensure that the most efficient combination of plant components is used to match the current cooling load demand.

Plant optimisation should preferably have an AI-based approach that is flexible and adaptive and utilises deep learning algorithms to shift with varying plant performance conditions. This is in contrast to the typical approach, where a plant technician would reset chilled water (CHW) supply temperature as the load increases, reset the condenser water (CW) temperature as wet bulb decreases, maintain a fixed Delta T between CHW leaving and CHW entering temperatures or control the speed according to a pre-set algorithm.

The plant optimiser should be able to adapt automatically to changing environmental conditions and system changes over time and analyse total system energy usage, including all plant room equipment and air-side energy consumption. It should be capable of controlling advanced functions, such as free cooling, heat recovery and thermal storage. It should also be able to provide a comprehensive dashboard view, detailed performance reporting and proactive action to address operational deviations.

An intelligent system should also employ Equal Marginal Performance Principle (EMPP) and Relational Control. EMPP is understood as the following: The energy performance of a system with multiple components is optimised when the change in system output per unit of energy is the same for all individual components. Relational Control is: System level control based on optimal power relationships between components. The Relational Control strategy also ensures that individual variable-speed devices, such as fans, pumps and chillers operate as closely as possible to their natural curves.

All these functions will then allow for greater plant efficiencies.

To summarise, the chiller plant optimisation system should have the following features and operating philosophy as minimum:

  • Full optimisation with comprehensive real-time reporting capabilities
  • Measure, verify and manage system performance in real-time
  • HVAC systems operate more efficiently when optimised at the system level – that is, running more equipment at part load
    • Operate chillers, fans, pumps and motors at varying speeds, based on power relationships and the demand placed on the system
    • Affinity Law’s impact on an all-variable-speed chiller plant; the law states that the energy used is proportional to the cube of the change in speed of the motor
    • For example, two pumps operating at 50% capacity use 75% less power than one pump operating at full capacity2
  • Detect, diagnose and correct system faults as they occur
  • Relational control
  • Feedback loop-based control and proportional/integral/derivative algorithms and adaptive tuning loops
  • Receive real-time data about system performance and building loads from the BAS
  • Calculate most energy-efficient operating sequences of the entire HVAC system algorithms
  • Feed information back to BAS to adjust system operation
  • Repeat optimisation process, say every 30-60 seconds

For what it’s worth

A traditional chiller plant manager system does not offer the level of optimisation that you would expect if you have very high-efficiency equipment, such as oil-free (perhaps magnetic-bearing) chillers with variable-frequency drive (VFD) that provide maximum efficiency at part-load conditions, and pumps and fans with VFDs, and are targeting around 0.6 kW/ton.

Having a traditional approach to increase the efficiencies of such chiller plants will be similar to flying an F-35 fighter jet without avionics and manually trying to lock on to several targets simultaneously.

What will fulfil this requirement, and more, is an intelligent optimisation system using AI algorithm built into its core that will address the earlier-stated requirements; facilitating continuous, dynamic optimisation and, hence, being able to deliver greater plant efficiencies that are in the excellent range. Optimising performance of the entire plant through enhanced management and improved software is where the real potential lies 3 … as there’s a point where you can make a piece of equipment only so efficient. And this is the very reason we will see the integration of AI into chillers and other major plant equipment sooner than later.

References:

  1. All Variable Speed Chiller Plants, ASHRAE Journal, September 2001
  2. Johnson Controls® Central Plant Optimization™
  3. https://www.achrnews.com/articles/143164-the-promise-of-artificial-intelligence-in-chillers-and-rooftops

 

Rehan Shahid is Director, P&T Architects & Engineers. He may be contacted at rehan@ptdubai.ae.


Share this story

Feedback for this story

Your email address will not be published. Required fields are marked *

FaceArmor Masks
Honeywell
Banner – Bitzer (17jan19)
Banner – Midea (12-17)
Banner - CCGD
Banner - AHRI
Banner – Matrix AVE
Banner – CareersBay