Accelerate data-driven circular economy initiatives with AWS

Jenna Lia: Welcome to this session on accelerating data driven circular economy initiatives with AWS. My name is Jenna Lia and I am part of the AWS Sustainability team, predominantly working with customers across our Asia Pacific region. I'm gonna be joined in the conversation a bit later on by Mr. Kam O who leads the AI Innovation team at SK Ecoplant, who's an AWS customer.

Today, we're gonna start with the fundamentals. What do we mean by circularity? What does it mean in the broader context of sustainability and why is transitioning to a circular economy important? Then we're gonna look at how organizations can get started on the transition by first adopting circular economy principles and we'll take a look at those being adopted by AWS. Finally, I'll consider that we need to be moving through the transition at pace and how digital technologies can help accelerate the outcomes of circular economy initiatives.

I am then gonna welcome Mr. Kom O. He will introduce SK Ecoplant and take you through an example of where his team has leveraged technology to accelerate initiatives in their own operations.

So for a while now, the terms sustainability and circularity have been gaining momentum in more mainstream political and social dialogue. Circularity is not a new term but the conversation on how to transition to a circular economy has now entered corporate boardrooms and government strategies around the world. Where an organization is striving to be more sustainable, they are more than likely trying to make their operations more circular as well.

The UN has defined sustainability to mean meeting the needs of the present without compromising the ability for future generations to meet their own needs. In order to do this, you have to acknowledge that sustainability takes a systems level approach. It considers the interrelationships between environmental, social and economic factors, how they interact and how they can be balanced.

You can't just focus on environmental aspects. For example, this may shift risks elsewhere in the system resulting in unsustainable outcomes. Instead, you must strive to understand all aspects and their impacts and ensure that they are balanced.

For example, think about a toothbrush where the handle is made from wood instead of plastic. It seems sustainable. But you really need to consider the often unqualified and unaccounted for sustainability related impacts that span the entire process that led that toothbrush to your hand before you can call it sustainable. That is to say that you need to look at all of the environmental, social and economic factors at play.

Then the concept of circularity - when a product or a resource is renewed or regenerated, rather than wasted. When a resource is in a circular economy, it is subsequently reused in a way that allows it to go back into a new life cycle or supply chain.

Think about that same toothbrush with the wooden handle. If that handle has been made by timber offcuts, that material is being reused in a new supply chain instead of being discarded as waste. Understanding these two concepts of circularity and sustainability, you can see that achieving circularity on its own does not necessarily mean something is sustainable, but that circularity is one method of achieving sustainability.

So at Amazon, we are committed to being net zero across our business by 2040. And circularity plays an important part in helping us achieve this goal. It is a priority across all areas of the business including AWS data centers. We want to ensure circularity is embedded in our operations because we know it is one way that will help us build a more sustainable business.

One example of where we're doing this in AWS data centers is using cleaner fuel for our backup generators. Starting this year, we have implemented hydro treated vegetable oil or HVO in data centers in Ireland and Sweden. HVO is a renewable, nontoxic fuel that can be made from vegetable or plant oil. Given this, HVO is apt for use in a circular economy.

Also, according to one of our fuel suppliers, using HVO is expected to reduce the CO2 emissions by up to 90% over the fuel's life cycle when compared with diesel from fossil fuels. So in the future, we aim to use HVO across all of our data center sites in Europe, and we're looking to expand that globally.

Another example is with our water usage in our data centers. Last year here at re:Invent, we talked to you about what it means to be water positive by 2030 - that is that we are going to return more water to the communities than what we consume in our direct operations.

Water is the single most important shared resource across all aspects of our lives and it's not only a resource, it's a nutrient carrier and it's even a source of energy. The longstanding approach of using water and then disposing of it straight away is no longer viable. Too many urban centers around the world are vulnerable to declining and variable water levels for us not to do anything.

So at AWS, to become water positive, we prioritize sustainable water sources such as recycled water wherever possible. In fact, recycled water is used for cooling in 20 of our AWS data centers around the world. This includes data centers in California, Virginia and Singapore where these data centers are using recycled wastewater instead of drinking water to make sure that higher quality water is available for the communities.

Also, once we've maximized the use of water in our data centers, we find ways to return it to the community. In Oregon for example, 96% of the cooling water from our data centers is supplied to local farmers who then use that to grow crops like corn. This is circularity, taking something that's left over and regenerating it back into nature via the farmers.

According to a 2019 European study, each year 90 billion tons of primary materials is extracted and used, but only 9% is then recycled. What is more worrying is that more recent reports suggest that the situation is only getting worse and that percentage is slipping back to 7%. This means the majority of our global economy is a linear one where raw materials are taken and used to make products. Then the products or its materials go straight into landfill and the process starts back at the beginning.

Once again, instead we need to be transitioning to a circular economy - one where we start to see the reduction of finite materials and waste created across end-to-end supply chains. So to do so, we need to be reversing linear trends and moving away from the current take-make-waste mentality that is sapping natural resources and dangerously warming the planet.

But it does go further than just recycling. In a circular economy, for example, the product is designed to use recycled content and then once the user is finished with the product, its components are returned back into the supply chain instead of going back into landfill.

And we can all do this. Right, hands up - who has kids? Keep your hands up if you're happy to be here at re:Invent without them. I'm kidding, hands down.

When I'm not working with AWS customers or here talking to you at re:Invent, I am a mom to my young son. Now I'm learning kids, they grow up fast and I'm constantly trying to think of new ways to engage him and give him space to grow. So I often find myself in scrap yards, finding material that is no longer in use and rightly or wrongly, I build toys and furniture for him to use.

Please don't judge the quality - to my son's dad this is how we're saving the family money. But this is a really simple example of how I'm bringing circularity into my own home using materials that would otherwise have been wasted. When my son is done with the toy or the shelves that I've built, I can then pass that along to a new home, keeping that material out of landfill once again.

So as individuals, we can all do our part, but we need to build circular operations in business. This is where the opportunity to transition is at scale and it allows us at a faster pace to decouple economic activity from the consumption of finite resources.

Look at how the World Economic Forum has defined circular economy - it will not come around by accident. There needs to be intent and the intent needs to start at the design phase.

So if we think about what a circular economy is, then it's really based on three principles:

1. Design out waste
2. Keep products and materials in use for as long as possible
3. Regenerate natural systems

Adopting these three principles is really the first step that any organization can take on the transition. And at AWS, we embrace these three circular economy principles for our server racks.

The first principle is design. In this stage, we eliminate excess materials such as steel or plastic, increase recycled and bio content, and plan for repair, reuse and recycling from the beginning.

Then we want to ensure our equipment can be used as long as possible. In 2022, AWS actually extended the life of our server racks from four years to five and our networking equipment from five years to six. In addition, we have a robust maintenance and repair program designed to increase component reuse and further reduce the waste across our supply chain.

There is a tradeoff in this step - we want to extend the useful life of our hardware, but at the same time, balance any efficiency loss by not replacing that hardware with newer, more energy efficient chips. We strike that right balance by being focused on using equipment for as long as operationally efficient.

The final principle is recover. When it's time for our server racks to be decommissioned, sanitized equipment is routed to our reverse logistics hubs that AWS uses.

So I'm going to play a short video for you now that allows you to have a look a little bit more behind the scenes at those global reverse logistics programs.

[video plays]

Right now, we are in our failure analysis testing lab which is a part of our AWS reverse logistics processes for North America. In the failure analysis lab, we evaluate repair and test components from retired server racks. All of the hardware that it takes to power our data centers is now able to come to our AWS reverse logistics hubs to be able to be repaired and reused.

Our goal is to extend the life of our existing hardware whenever possible. When the hardware arrives, it's in AWS racks, we apply really the same high level of care and data security as we do when the hardware is active in production in our data centers. All of these server racks we receive are sanitized of customer data so they can be remanufactured during this process.

The servers are disassembled so that we can access the components, we might be able to repair and reuse and then we will test everything from switches to supply units, DRAMs, graphics processing units and then optics for our Nitro cards.

After they go through the de-manufacturing process, they are eligible for our failure analysis testing process. When a card gets to the FA lab, the first thing we do is a visual inspection. We do have a repair process that allows us to repair some of the components on the card to allow us to further test the card.

We put the repaired cards in testing rigs that simulate the environment of a data center. And then we have a series of software scripts that work in conjunction with the hardware to determine if the card will perform to the same high quality standards.

Jenna Lia: So in summary, adopting circular economy principles is an important first step for any organization looking to operate more sustainably. At AWS, we're committed to building circularity across our operations from the way we design our server racks to how we manage them at end of life. But there is an urgency required here. We need to accelerate the transition.

Digital technologies like IoT sensors, data analytics, and machine learning can help organizations implement and then scale circular economy initiatives faster. We need to leverage these technologies to help drive the change at pace.

I'm now going to hand over to Mr. Kam O from SK Ecoplant who's going to talk through an example of how they have used technology to accelerate circularity outcomes.

Kam O: [Gives presentation]

Jenna Lia: Thank you Kam for sharing the great work SK Ecoplant is doing. It's been fantastic to have you join today. So in closing, transitioning to a circular economy is critical and it's a key enabler for organizations to meet their own sustainability goals and commitments. By adopting circular economy principles and then leveraging technologies like the ones Kam shared, organizations can accelerate their progress.

Just as a new card from manufacturing, all of our testing process is supported by Amazon's Annapurna Labs teams to determine what needs changed, adjusted or modified into the new card manufacturing. As a final step, the functioning cards are returned to inventory so they can be reused in one of our data centers. This process would not be possible without all of the people behind the scenes working together to solve these really interesting challenges that have not been solved on this scale before.

The work that we do here is creating a more sustainable future for our data centers. And we're really proud to be a part of that. We have a lot of pride in the work that we do and knowing that there's still a lot of useful life in the products that we touch to support all of the services that our AWS data centers support for customers.

So in that example, there was a piece about providing feedback to the team that originally designed the equipment. And I think that's an example of where digital technologies can help to accelerate a step like that when it is at scale by allowing teams that are running these type of initiatives to execute them using near more near real time data, we know the window to act and prevent the devastating impacts of climate change is closing rapidly. So finding ways like this to accelerate these initiatives through the use of technology is crucial.

But when I'm out in the field, talking to customers and especially IT teams, they balk at the the topic of sustainability and even more so circularity. But I want each of you in this room to consider these topics to be just another business use case. And I would guess all of you in this room with your expertise in working with AWS technologies can help to accelerate outcomes. You can support your operations to be more circular and therefore your organizations to be more sustainable through the use of our technology.

So let's look at some examples of this that is how digital technologies can accelerate circular economy initiatives. First up, IoT - the extension of the digital world into the physical. IoT is a network of physical objects linked by sensors, software and the internet for the purpose of exchanging data. Sensors can collect data about the behavior of processes like conditions such as temperature or moisture. There are so many potential use cases in this space given that machine to machine communications from IoT devices can generate more data than human to human communications.

IoT technology is crucial in a circular system because it enables the creation and the processing of data to give organizations better viability into real-time material flows and their supply chain. Then what do you do with the data? Data can help to establish baselines for sustainability use cases. And we are seeing many customers do this for carbon, for example, as part of their decarbonization goals.

However, thinking about baselining when it comes to transitioning to a circular economy is nascent as is assessing levers and measuring outcomes which is still patchy and not comprehensively understood. So foundational data projects could help to identify and close a lot of these gaps by bringing silo data into a circular economy data lake. Then organizations can use that to establish baseline and set goals to make incremental progress or improvements against it.

Data can also be a powerful insight as to trends roadblocks and enable organizations to track KPIs as to the effectiveness of their initiatives. Implemented examples of KPIs in this space could be the amount of waste generated through the process or you could assess how much water you had planned to use versus how much you are actually using in downstream processes.

The last one that I want to touch on is AI/ML. These technologies can help to identify patterns and plug data gaps as well. Or together with predictive analytics and simulations, it can forecast future changes to that established baseline that you are able to make. AI and ML algorithms can monitor and optimize energy use in real time, identify inefficient processes and recommend changes to reduce waste.

For example, we can use machine learning to analyze trends and detect patterns that lead to better decision making. IoT when combined with AI could allow for better monitoring and management of resources. According to the UN Environmental Program, data analysis for innovation, forecasting and optimization is the most significant application of AI when it comes to circular economy.

So one example of how these technologies are actually working in practice is with an AWS customer, Reseau Car, one of the largest buyers and sellers of used vehicles in France. Using a combination of real world and synthetic data, the customer built an ML algorithm using Amazon SageMaker to analyze images and determine the cost of refurbishment with exceptional accuracy by accurately detecting vehicle damage and assessing the refurbishment needs on AWS.

Reseau Car's trade in processes are faster and more efficient. This ability actually extended the lifespan of selected cars by up to five years which helped reduce landfill waste and promoted sustainability within the automotive industry.

So at AWS, we are also using digital technologies to accelerate outcomes of our own initiatives. One example is improving our water efficiency. So in the data center space, we need water to cool our data centers. And when doing so, we strive to use that water in the most efficient way. In 2022 AWS's global water use efficiency metric was 0.19L per kilowatt hour. And that was actually a 24% improvement from 2021. And it's AWS technologies that our teams have been able to leverage, that drove some of this improvement.

One example is how our water teams have used simulated expected water use and compared that to the measurement of actual water use in the data centers. The expected use in the past has come from data center designs, but the team was able to incorporate weather data from a third party source NOAA to improve the accuracy of expected water use. What this then allowed was that they could respond to operational inefficiencies in a more timely manner and stop water leakages.

For example, given that 2% of the global water supply is suitable for human consumption, we really need to avoid it being wasted. The EPA estimates that 1.7 trillion gallons, roughly 30% of all treated water is wasted every year in the United States alone. For water utilities around the world, aging infrastructure is increasingly becoming a problem to actively manage the water system. Getting information like consumption, pressure and flow data in real time is crucial and it's architectures like this one I'm sharing which shows you how AWS IoT Core for LoRaWAN can be used to reliably collect water meter data from meter readings from multiple sources, multiple devices in the network, this would allow you to transfer the data to the cloud, detect water leakage in the grid and gain deeper insights into water consumption.

So leveraging digital technologies to accelerate circularity requires this combination of technology and data. Doing so can create a more sustainable and efficient way of doing business and this is what SK Ecoplant has done.

To tell you more, please join with me in welcoming Mr. Kim to the stage.

Yeah. Thank you for introducing me Jenna. I'm Kim, a member of SK Ecoplant AI innovation team. I'm really grateful for presenting today and engaging in discourse on the environmental issues. In this presentation, I will explain how SK Ecoplant is innovating the environmental industry and solving the environmental problems through digital technology and data.

First of all, I also like to briefly introduce SK Group and SK Ecoplant before delving into the main part. SK Group is one of the largest companies in South Korea like Samsung and Hyundai and the revenue is $138 billion. Also, SK Group has over 180 subsidiaries. In 2023 SK Group was ranked as the 9th largest company in the world by Fortune. SK Group focuses on four key areas: semiconductors, energy, telecom and bio. Among these SK Ecoplant where I work play a pivotal role in the green sector.

If you think about the circular economy, it basically means that waste is not discarded, but rather returned to the energy or resources. If you take a closer look at the entire waste life cycle, we can see that many companies are discussing the concept of a circular economy. However, it is important to note these companies are still managing waste in a traditional way without utilizing any advanced technologies. The environmental industry still has a lot of work done by manual process and phone communication. Simply there's no integrated system. Companies in this area also concentrate on each other's work. So that makes their businesses fragmented and suboptimal. Potentially challenging to integrate with traditional ways.

In this situation, let me run you through how SK Ecoplant is innovating the environmental problem to circular economy, data and connections. We are focusing on two keywords to create a connection between fragmented environmental business. SK Ecoplant collects and analyzes data for it. I will explain more about our plan.

Now, you can see our company in the middle. Companies play an important role in our society, but unfortunately, they are also a major problem related to carbon and waste. Our approach is measuring the amount of carbon and waste emissions by companies. On the left side, you will see a concept of our carbon management named the Weibo - the carbon. The first step is measuring the amount of carbon. If these companies engage in activities that offset their carbon emissions, such as trading carbon credits, our platform will give them credits for their efforts.

On the right side is the concept of our circular economy. SK Ecoplant developed its own digital waste management platform named Circular Weibo. It won the CES innovation award in 2023 and it can help digitalize and manage the process of returning waste. Therefore, we can obtain data via Weibo that provides value of insights into what needs to be solved and how to solve it. And it doesn't just end with that.

We developed We Energy for incinerators and We Reward for wastewater treatment. First, I will briefly introduce We Energy and We Circular and then I will explain We Reward in detail.

We Energy is a solution that increases efficiency and reduce emissions of existing engines. It has been applied to five incinerators in South Korea region - 12% max reduction and 3% energy recovery rates increase. It is currently being applied in Southeast Asia as well.

We Circular is a digital waste management platform. It provides visibility and transparency and it is currently used by over 200 companies and has reduced efforts by over 50% by simplifying the complex process.

Now let me provide you with background information of the wastewater treatment process. It will be better for your understanding. In brief, after the wastewater enters, it goes to the settling tank for sedimentation. Then it undergoes biological treatment in bio reactor followed by chemical treatment in the disinfection process. Finally, the water is discharged. The treatment process itself is not very complex but dealing with the water is still one of the most difficult fields due to various regions such as the long rotation time and difficulty in measuring the exact reaction.

Also, there were more challenges than we initially thought. First, only the final water quality was being measured for regulation and there was not much data being collected during the intermediate stages. Second, as there are no measurement and no prediction, its operation relied on the operator's experience.

Moreover, since there was no prediction, it safe chemicals and biological energy were being used for the final discharge research.

Third, even if sensors were installed, uh they were easily contaminated as wastewater continues to flow, then they needed to be cleaned and calibrated continuously. Usually this process uh needed to be done every uh it needed to be done once every two months. But those who manage it were busy and sometimes sensor cleaning was neglected uh visiting in in a data collection. Naturally, prediction that were forecasted using ino data lost the trust of the operators and they eventually returned to the to the past operation.

We solved these challenges with AWS. We have equipped various sensors to enable real time monitoring and developed a digital management platform for it. Next, we predict the affluent world quality based on the data uh rather than the operator's experience so that they could operate optimally. Finally, we developed a i basis of their calibration to solve sensor contamination.

Here is a service screen of we layer that i explained earlier. It consists of three main parts.

Firstly, the blue area is a i guidance for air flow which guides the amount of air injected for microorganisms.

The yellow area is a i guidance for chemical flow which guides the amount of chemicals to remove tiny particles.

Lastly, the green area is a i based water quality prediction which forecast the water quality nine hours later, the a a i guidance has reduced operating costs. While world what model quality prediction can forecast future world quality so that it makes the operators can work optimally in advance.

In particular water quality prediction shows performance of over 85% increase by one minus ma p criteria. I believe that it is top level across in this area. Is it not this is it is a i based software calibration. As as i mentioned earlier, the sensors are easily contaminated by waste water making it difficult to keep measuring a great data.

Due to this situation, operators do not trust the sensor data and change their operating operating conditions. Using the web data that is only sampled once a day a i based soft calibration is a service that collects contaminated data values using a i. When the sensor data differs significantly from the lab data i just shown in the graph below. The orange line is the low data from sensors and green data represents the web data. It is a kind of ground truth when the sense is virtually calibrated by our a i. The data is collected as shown by the blue line and it shows the distribution similar to the lab data.

Through this, the reliability of the sensor can be maintained for a long time and a maintenance period can be extended. So reducing the workload for a physical calibration.

Now, i would like to explain how we implemented the weibo with AWS. Let me tell you more technical details.

Sk eco plants operates over 1000 water treatment plants in South Korea. In the near future, we are preparing introduce we reward for achieving centralized counter management. I also like to explain what AWS services we used and what difficulties we overcame.

The first was the CDK service. There were time and financial difficulties to deploy weibo re from the outset utilizing CDK allowed us to simply modify the solution to fit each place unique characteristics.

The second was Kash service by utilizing KS or streaming data pipeline could be easily and quickly configured based on the data size.

The third was the ECR service. By utilizing ECR, the model could be easily managed by version and if it could be easily logged back to the previous one.

The fourth one is SageMaker, AWS SageMaker have to choose the appropriate algorithms for each plans and provide various tools for easy model deployment.

Next, let me explain the architecture of Weibo newer. The field and weather data are collected via Kinesis. When examining the area of where the collective data is stored, the data is loaded into S3 gold S3 low also is then resampled and restored in S3 gold. Our AI model is that trained and deployed using the S3 gold. Also, the code is stored by Comets in Code Comet looking at the model influence process. AWS Lambda is used uh lambda brings our real time data to the VPC and store data in Influx DB. Then using the model, it infers the water quality prediction values.

Now let's take a closer look at some of the key AWS services.

The first one is CI/CD data pipeline. Weibo le reused many parts from the previously developed Weibo reen energy that i simply introduced earlier. Uh this is easily made possible by utilizing CDK. We plan to expand the solution to more sites by 2024. We expected to be able to easily configure new sites and make models using AWS called pipeline. When configuring CDK and Code Pipeline. The code is modified to reflect the characteristics of each site depending on characteristics. The modified code is managed by theory uh in Code Come for example, when training a model in Stage Maker, it manages the optimal pipe parameters found for each site.

Next i will talk about Kinesis data stream. About 5000 types of data are collected from each site. And among these 900 types are utilized in Weibo r via Kinesis. As we expands to more sites in the future, more real time data processing will be required. Kinesis can easily and quickly process time sheet data without losing data. It is essential for Weibo.

The order and the most important thing is SageMaker for water quality prediction. Weibo r has a total of nine models for each model is simultaneously trained using SageMaker whenever retraining is required. In addition, when introducing new algorithms and hyper parameter optimization is also performed concurrently using the Maker as i mentioned earlier with oral treatment undergo biological fortification which utilize our micro organisms, micro organisms are highly sensitive to seasonality due to the impact of their environment. Weibo reorder utilizes ECR to automatically manage the version of our models. And if a model is trained that props poy for any number of regions such as sensor contamination and environmental changes, it can be easily loaded back to the previous version,

Weibo we was developed over one year. It took a lot of time to get an AI model that reflects the seasonality rather than sim rather than simply setting up the system and installing the sensors. Many operators said that AI guidance and world quality prediction could provide support for optimum operation and based on this, it could save resources.

The graph on this slide is comparison of the broad power consumption during the, during the same period last year. And since we, we order was applied, we have reduced the power rates by 12% compared to the previous year, which can lead to 19 19% reduction in power consumption when convert them to greenhouse gas. This is the same effect as reducing GHG emissions by 13%. It is simply possible to calculate well on formula.

Additionally, all do not discussed today. We have our achieved 10% reduction in usage of chemicals through AI guidance. When uh the automation phase using self driving cars is applied to waste or treatment. Most existing plants are in one or two phase by applying our solution. They are able to operate at level three to achieve natural and circular economy.

We believe that they need to be controlled in a more advanced way to reduce carbon and energy consumption. For this, more digital technology is needed and SK plants will continue to develop We to secure sustainable pro processing technology.

SK eco plant is trying to solve many environmental problems through digital technology and data. And as i mentioned today, we have achieved huge progress in this environmental businesses. If you are interested in our solution and technology, please visit our website and contact me.

Thank you for listening. Now, i'll hand back to J

What a great example. I, i think it's when i first came across it and mr kim, oh i was so excited cos it really embodies what we've spoken about today, which is what is circularity and why circularity and circular economy and transitioning to one is important and we need to accelerate it and the way that they have done that through technology and data, i, i was really impressed and i, i hope you have enjoyed it today. As i said earlier, a lot of you in this room or watching online have the skills to do similar things within your own organization. So i'm going to give you all a homework assignment, go back into your organizations and find opportunities where AWS technologies could accelerate existing circular economy initiatives or at least start the conversation on how processes that you're aware of can be moved to be less linear and more circular and then how technologies could accelerate those outcomes.

Also be sure to check out our other sustainability sessions scheduled this week. And there are more uh that you're able to watch on demand that happened earlier in the week. And myself and my colleagues, we are all at uh Caesars Forum at the Sustainability Showcase both today. And tomorrow. So please, if you're around stop by. We have a lot of other examples of where technologies and data have come together uh to either result in sustainable outcomes or accelerate some initiatives in in motion.

So, on behalf of Mr Kao and myself, i uh i wanna thank you for being part of the conversation. I hope you learned something today about circularity and why it is such an integral step to take to building a more sustainable business and a more sustainable world.

And so speaking of being data driven, we wanna capture your opinion. So please take a moment uh to complete the session survey in the mobile app.

And thank you very much and see you next time.