REDEFINING DESALINATION WITH AI

Energy efficiency and life-cycle cost optimisation are among the most important challenges for utilities and for developers within the context of desalination plants. Today, Artificial Intelligence (AI)

and machine learning are helping operators make real-time decisions to save time and money

Over the past half-century, there have been tremendous advancements in the efficiency and scale of conventional desalination technologies. Today, there are more than 20,000

28 global desalination plants in operation, providing more than 28 billion gallons per day of potable drinking water in 150 different countries, according to the International Desalination Association (IDA).

Yet, fresh drinking water remains one of Earth’s most precious commodities. Although conservation efforts and policy initiatives are critical, technological advancements will be needed to ensure that supply

keeps up with demand.

The dominant desalination technology is reverse osmosis (RO) membrane technology, which accounts for 60% of the global capacity, and is expected to grow in the coming years, according to the International Renewable Energy Agency’s “Water Desalination Using Renewable Energy: Technology Brief.”

This increase is due to significant improvements in RO membranes, pre-treatment and energy recovery, which have decreased RO desalination cost.

However, cost reductions have begun to plateau, and the process is still dogged by high-energy consumption and performance instability caused by sensitivity to variations in feed-water quality.

The industry has reached a point where advancements in the economics of complex computational processing are providing real solutions. Artificial intelligence (AI), or machine learning, is the future of the water industry.

Semi-autonomous computers are everywhere: Autopilot is making air travel safer and faster; mobile phones are like pocketsize personal assistants; and self-driving cars are promising to increase safety in daily commutes.

Yet, the operation of vital public infrastructure—water and wastewater treatment plants—is still reliant on Victorian-era technologies. Plant operators may be seen standing in front of a water treatment system, waiting for something to go wrong.

Despite the technologically advanced age, water treatment operators and engineers remain the first line of defence to combat complex, natural and dynamic system disruptions. Even well-trained, experienced plant personnel require days to process data, analyse trends and prescribe changes needed at any given time.

Such delays in response can lead to overstress and reduced lifetime of process equipment (e.g., pumps, membranes, etc.), lowered throughput, and the overuse of costly consumables (e.g., cleaning chemicals, antiscalants, etc.).

However, AI can play a pivotal role in making society’s current desalination infrastructure more cost-effective, energy efficient and, ultimately, better equipped to selfadapt and self-optimize to the inevitable variability of process conditions.

The advanced mathematics used to optimally maintain a complex water treatment system cannot be processed by humans in real-time.

On the other hand, computers excel at rapid computing—around the clock and with perfect memory. Advanced computing and control philosophies will allow operators, engineers and their companies to make more informed decisions in a timely manner.

AI technology can be used to improve the desalination process by optimizing supplemental equipment surrounding the desalination membrane. One self-adaptive flux enhancement and recovery control technology appears to be the first application of AI as an active, real-time control platform in the water industry.

The technology monitors key operating parameters, performs real-time data analytics and makes predictions about when and what future maintenance actions will be required for upstream ultrafiltration pretreatment systems, which act as primary barriers protecting downstream RO membranes.

In doing so, it provides the ability for the pre-treatment filter to adapt to high fouling events buffering against the effects of these events on downstream membranes—maximizing plant uptime and potable water production.

Integrating AI into the water industry will maximize the potential of current technology while freeing valuable time for experts to focus on these higher-level advancements.

The physical membrane material has complicated properties that are not observable, and the RO process itself has pieces that cannot be measured or approximated mathematically. To optimize RO, where standard mathematics has limitations, machine learning is accurate.

Machine learning can also be codified and deployed to systems like SCADA to predict variations/trends in water temperature and salinity and undertake multiple set point changes per day, ultimately minimizing energy use and adapting to consistently fluctuating feedwater conditions.

Software can be designed to account for multiple input and output parameters, ensuring each plant’s design standards and mechanical limitations are met.

When optimising multiple trains, it’s impossible for humans to calculate the optimal flow balance at a frequency to deliver the most energy savings. For example, if a person used Microsoft Excel to do the same calculations, it could take a whole week to calculate savings for one train, so you can imagine the complexity of 20 or more.

Multi-train optimisation can also simulate multiple trains of RO racks, while approximating the properties of the plant that are difficult to model mathematically. The simulation, combined with machine learning, finds states that reflect the best ways to meet KPIs of a particular plant. This is an exciting area for desalination, which can add further efficiency than industry-standard operations.

Water utilities are starting to realise the need for integration of AI and a lot of investments are currently being channelled into that direction.

29

In May, Aquatech International, a global desalination leader, partnered with Pani Energy, an artificial intelligence (AI) analytics solution provider for water applications, to reduce the energy and cost associated with desalination.

The partnership aims to augment Aquatech’s proprietary LoWatt® membrane desalination process with Pani’s artificial intelligence platform to reduce the energy required for desalination to levels that have not been achievable using existing approaches.

LoWatt is a low energy desalination process that integrates robust pretreatment, optimized reverse osmosis design, and a proprietary cleaning mechanism to deliver what is currently the lowest specific energy consumption of reverse osmosis-based desalination.

LoWatt’s performance already betters the current industry standard and offers additional advantages including higher reliability and better plant uptime, thus reducing the overall cost of desalination. The integration of Pani’s software solution will help Aquatech further lower the energy consumption and establish a new industry standard of 2.7 kWh

30 per cubic meter (m3).

Enabling this lower energy consumption is expected to increase adoption of desalination in water scarce regions.

“By leveraging existing data, systems, and people paired with machine learning techniques, current desalination technologies can be lifted to new standards of efficiency – making water less expensive to produce and delivering affordable, climate resilient water to people and communities in water stressed regions,” says Pani CEO, Devesh Bharadwaj.

“With superior process design combined with advanced machine learning, Aquatech can provide a solution that reliably meets treatment goals while minimizing energy consumption and O&M requirements in realtime.

Ravi Chidambaran, Aquatech’s Chief Operating Officer says that the solution will help meet treatment goals while minimising energy consumption and O&M requirements in real-time.

“This partnership us to better serve our customers and address the biggest pain points of desalination – energy consumption and biofouling,” says Chidambaran.

In March, global water infrastructure developer, ACCIONA, said it would finance an Artificial Intelligence project at the Umm Al Houl desalination plant in Qatar from its dedicated decarbonization fund – one of only 14 projects in the world that have been chosen for their potential to significantly reduce carbon emissions associated with ACCIONA’s business activities.

The Umm Al Houl Seawater Reverse Osmosis plant, which is scheduled to become operational in April, will use a state-of-the-art AI platform called Maestro, to optimize operations and achieve energy saving in particular.

ACCIONA will use AI to lower energy and reagent consumption, which will lower the carbon footprint of the desalination plant. ACCIONA estimates the use of AI will reduce emissions by 12,000 tons of CO2 per year.

The Maestro AI platform processes operational data in real-time to allow predictive, autonomous and continuous optimisation at scale. This is expected to deliver lower operational costs, while maximising output, plant reliability and water quality.

The Maestro platform is also scaleable, which will allow ACCIONA to extend the

benefits of AI to all its clients in the global water industry as part of its commitment to continuous innovation and service excellence.

In Western Australia, machine learning deployed by Synauta at a 4,000m3/day SWRO plant realised up to 18% instantaneous energy savings and an average of 9.7% energy savings over six months.

The plant is operated by Osmoflo and has a typical seawater RO arrangement, with an isobaric energy recovery device and 4 trains capable of producing 1,000m³/day.

The trains have 14 vessels and 6 membrane elements in each. Elements are 6,000 GPD flux and 99.7% rejection sourced from a major RO membrane manufacturer.

The key to innovation, especially in the early stages, is to keep it simple and keep it focused. In this case, to save energy, the software works by manipulating plant recovery based on three set points. Operators simply enter the recommendations rather than undertaking lengthy calculations, which most operators do not have the time to do each day.

While solutions like using renewables and Energy Recovery Inc. are addressing energy needs, desalination can realize further optimization through machine learning.

By optimizing RO to save energy costs alone, the OPEX savings through machine learning could be more than $3.5 million per annum for a mega-plant like Al Taweelah IWP. With consistent pressure to lower water prices, and costs now under $0.30/m3, machine learning optimisation is another tool to help the pricing challenge.

Reducing energy and chemical consumption in desalination also saves CO2e, which is critical to mitigate climate change. Purely optimising RO to save energy, it is estimated that machine learning can save up to 12 million tonnes of CO2e per year across our industry.

Optimization of design doesn’t stop with capex savings. Optimizing performance capability and reducing losses will also bring energy efficiency with every kW/h not being used, creating opex savings. The combination of optimized products and systems provides positive impact on energy usage and cost.

Increasing the efficiency of the process makes potable water more affordable. Energy consumption can be reduced through economical design. High-performance equipment requires less energy for the same output. High efficiency motors use less energy to produce the same power, thereby reducing energy consumption rates and bringing cost efficiencies.

There are several digital optimising products available to the water industry. Consider the membrane, the heart of a water desalination plant, for example: digital solutions can be used to develop a model of the membrane to collect data and analyse it in different conditions, with different operational factors to predict the optimum time to regenerate.

Digitalisation enables the replacement of the membrane to filter the desalinated water and produce potable water based on its condition rather than at a regular frequency. The aim is to neither under maintain nor over maintain.

31

However, you need the correct data to avoid replacing too early (which brings with it the unwanted cost and impact of unnecessary waste and shutdowns) or replacing too late and consuming more energy and impairing the quality of the water. This can occur when the membrane has been used too long, requiring additional chemicals to convert the desalinated water into potable drinking water.

As we can see, there are consequences of not using digitalization and predictive models to maintain water plants at the optimum time. The benefits are less waste, less energy consumption, and the use of fewer chemicals. Additionally, shutdowns can be scheduled at an optimal time for operations, maximizing the life of the membrane and minimizing the use of chemicals.

Digitalisation enables water customers to use data to make decisions at the enterprise level that were formerly only possible at plant level. ABB, for example, has experts at its collaboration centres who collect and monitor data. These experts can then analyse and inform operations about every critical asset in a plant from the pump, transformer, and drive to the membrane.

Machine learning, artificial intelligence (AI), and algorithms are used to predict failure and the impact on the plant from both the unscheduled shutdown and damage.

This maximizes revenue and the resulting

32 operational availability contributes to improving the bottom line.

As well as enabling decision-making at an enterprise level, ensuring correct decisions are made at the plant level, with solutions available to help operators become better at making these decisions, is crucial to minimize costs.

In many plants, there are just too many alarm and warning systems — so many, in fact, that the alarms are frequently misunderstood, ignored or missed. Experts can redesign system alarm logics and algorithms so that operators only get actionable alarms to focus operations in a critical area of the plant.

Artificially produced water is key to life in many areas of the world where there are no natural water resources available. These plants are often located in hostile environments where there can be attempts to disrupt both the infrastructure and operations, impacting safety and security.

Thus, security systems can protect plants from intrusion and interruption. Further, remote access to diagnose issues can be a very cost-effective maintenance solution. It can be easier and quicker to access some plants virtually, rather than physically, without impacting operations.

Data can be collected to predict the condition of assets and decide the optimum time, interval, and conditions to optimize maintenance. Monitoring centres collect data from variable frequency drives and use the data to analyse trends, signals, and alarms, and identify potential issues to advise on optimising maintenance.

THE CHALLENGE OF OPTIMIZING ENERGY USAGE AND LIFE CYCLE COSTS

Energy efficiency and life-cycle cost optimization are among the most important challenges for utilities and for developers who are responsible for building and operating plants for several years, recovering their investments selling the water at agreed prices.

For organizational stakeholders, there are four key ways to make a positive impact on profitability — including reducing capex and opex as we have seen, and additionally taking steps to assure the lowest possible finance and possible risk to achieve a desired profit margin.

Although the desired profit margin is in the hands of developers and operators, working with a technology leader with experience in the water industry can be very beneficial, for many reasons.

In addition to helping to lower costs and maximize profitability, collaborating with an experienced, trusted partner can lead to improvements in quality and reliability, as well as consistency in service provisions. By reducing risks and lowering the costs of finance, water projects can become more viable.