Practical ways to apply data analytics and AI for energy management and emission control in the steel and cement industries

Disruptive technologies such as AI, machine learning and advanced data analytics – combined with deep domain knowledge and investment in a digitally literate workforce – are the key to improving sustainability and productivity in energy-intensive process industries, explains ABB’s Sanjit Shewale.

Share this page

Want to know more about AI?

Contact US
The digitalisation of the process industries – including in the chemical engineering space – continues apace, with innovations such as artificial intelligence (AI), machine learning (ML) and advanced data analytics offering unprecedented visualisation and pattern recognition capabilities, recommending optimal actions for every stage of the value chain.

Armed with these granular insights, industrial companies can make informed, data-driven decisions that drive genuine business value, reducing both OPEX and CAPEX, while improving safety, protecting their reputation, boosting environmental, social and governance (ESG) credentials, and complying with evolving regulations.

FIGURE 1: Energy-intensive process industries are evolving
Given these strong business imperatives, it is surprising therefore that many energy-saving and emission reduction opportunities remain untapped. Plant owners that adopt digital solutions can look forward to better performance at a reduced capital cost, as well as overall standardization of environmental practices, a significant help in reducing and managing their inherent complexity.

The issue of sustainability is now a priority, of course. In October, I had the pleasure of speaking at IChemE's ‘Advances in the Digitalisation of the Process Industries’ conference on practical ways to apply data analytics and AI to reduce environmental damage and create a more sustainable world.

The following month, more than 200 concerned parties, from national governments and businesses to NGOs1, gathered in Glasgow for the COP26 climate summit. The resulting Glasgow Climate Pact2 will accelerate action on climate change and efforts to limit the rise in global temperature to 1.5°C.

In this article, I will outline how AI-based digital solutions can be practically applied in energy-intensive process industries for energy management and emission control. Specifically, we will look at real-life use cases from the steel and cement plants, with short payback times and long-term benefits.

I will also touch upon the latest possibilities around using enterprise-grade platforms and suites of AI and ML based applications to establish a single source of truth system for the entire energy mix planning, consumption, and conversion, emission control and regulatory compliance, and the benefits this offers.
FIGURE 2: Examples of industrial analytics for environmental monitoring and compliance (GHG)

Plant owners that adopt digital solutions can look forward to better performance at a reduced capital cost, overall standardization of environmental practices, a significant help in reducing and managing their inherent complexity.

Technology is nothing without domain knowledge

In addition to the recurrent challenges of optimizing efficiency, and maintaining throughput and quality control during high-volume production, chemical processing companies must now comply with laws aimed at reducing emissions in line with the 2015 Paris Agreement – or face penalties.

The cement industry alone may be responsible for as much as 9% of global COâ‚‚ emissions. In the UK, that figure is 6% for industrial emissions. The Climate Change Committee has recommended setting a target of near-zero emissions for cement production by 2040 using techniques including carbon capture utilization and storage, thermal efficiency and reductions in the ratio of clinker to cement.

The whole energy ecosystem is also evolving – and while this transition is not without its challenges, the enabler is most definitely digital technologies. However, it is important to note at the outset that the idea that simply gathering data in a pool and then applying AI and ML will magically transform industrial operations is flawed. What is often missing is a conceptual understanding of how the data got into that pool – and, as importantly, how it can best be leveraged to add real value to a business.

Many technology/software companies dive into a project without having a clue about the industry. They make false promises. ML requires a lot of data – with data classification and management approach closely aligned to the business strategy. It also requires specialists able to literally teach machines how to structure their computing methods more like human thought, how to solve complex industry-specific tasks by breaking them into simple steps and how to self-monitor for effectiveness. Partnering with a trusted technology provider with deep domain knowledge and expertise is a prerequisite of success, and an incredible opportunity for companies and people who understand how to navigate the digital storm.

In short, the role of people in solving challenges around accuracy of AI techniques and how to explain them must not be underestimated – so that machine recommendations can be trusted. Data cleansing, anomaly removal, analysing the correlation of parameters and proper result interpretation are as important as ever; after all, if you trust the people, you trust the algorithm. I will talk more about investment in human capital later.

Case study 1: Decarbonizing steel production

When we talk about decarbonization potential across industries, it is really the additional ‘horsepower’ provided by advanced data analytics, AI and ML that is the gamechanger. For operators in the process industries, it will be impossible to hit KPIs and tightening sustainability targets by manual setpoint adjustment alone, and not just around reducing emissions, but also energy optimisation and management. As more and more data is there, and with the implementation effort decreasing, we will see more and more people getting involved in AI advisory - driven by corporate decarbonization initiatives.

Let’s take a real-life example of a steel plant with annual capacity of up to five million tons of steel.
FIGURE 3: Digital energy management can optimize steelmaking processes

We engineers know that it is easier to speak about ML than to implement it. We get excited about addressing the untapped opportunities in process industries, especially sustainability. But we also get practical and pragmatic, starting with examples of how data has already been used successfully.

Complex distribution networks for electricity, steam, by-product gases and imported fuels account for as much as 20% of production costs. Iron ore reduction is where the vast majority of carbon emissions come from in steelmaking, but the world will continue to rely on iron ore until around 2100. The challenge for the industry in the coming decades must therefore be to transition to alternative, preferably renewable, sources of energy and optimize their use for iron ore reduction.

ABB’s Digital Energy Management solution has a proven track record of short payback times and long-term benefits. It not only optimizes production scheduling, throughput and quality, but also energy-related costs, raw material usage, carbon, greenhouse gases and waste emissions. As mentioned, it relies on decades of experience in the process industries – and steelmaking in particular – captured in rule-based predictive energy management algorithms.

Designing, deploying and maintaining the optimal site-wide or enterprise-wide energy management and emission control strategy is a large engineering and operational challenge, one that requires a wide span of competencies, engineering tools, architecture approaches and service capabilities in order to identify the most performing and cost-effective solution for each process area and site.

Typical steps and modular approach for deploying a digital energy management system are:

Monitor: Monitoring energy usage at plant and process level with real-time visual displays and data
Identify: Identify best performance of process areas and opportunities for improvement
Report: Report energy consumption patterns of process areas and production lines
Analyze: Analyze inefficiencies in plant and process areas
Alarm: Alarm capabilities – enable corrective measures to be taken immediately
Forecast: Forecast energy consumption schedules for process areas based on production plans and measured consumption
Optimize: Solve economic real-time optimization problems consisting of own generation, trading and using of energy in industrial plants and power plants

FIGURE 4: Deploying digital energy management system in a modular approach
Waste gas utilization in metals production can be greatly improved by monitoring generation and consumption across plant facilities in real time. Data is collected from multiple systems to compare allocation with actual consumption, provide real-time demand and supply calculations, balancing, benchmark and optimal distribution, as well as forecasts based on production plans and historical data modelling. Root cause analysis is also applied when a gap occurs between supply and demand.
Figure 5: Byproduct gases network optimization

ABB’s Digital Energy Management optimizes production scheduling, throughput and quality, but also energy-related costs, raw material usage, carbon, greenhouse gases and waste emissions

Deploying data with edge computing power

Complementing these management systems with data contextualization and digital twins allows operators to further optimize their energy consumption using load profiling and balancing, and process parameters; air temperature at the compressor intake, for example.

As another example, a complementary System Anomaly detection app can be deployed to detect unusual pattern, anomalous behavior from process streaming time series data. It uses AI/ML methods to continuously monitor changes in pattern across IT and OT systems from high frequency near real-time data. By triggering alerts, it helps reduce manual efforts needed to identify and rectify energy consumption deviations.
FIGURE 6: System anomaly detection on asset or system energy consumption deviations and alerts
As I mentioned towards the beginning of this article, the latest enterprise-grade platforms and suites of AI and ML enabled applications enable companies to deploy and monitor advanced controllers, data analytics and optimization solutions at the edge, to and from an industrial cloud/multi-cloud, or on-premise.

However, this increased complexity requires digital solution providers to have the ability to master very different technologies, industry-specific processes, cybersecurity and offer the needed consultancy and assistance to the end users from the early design stages to system commissioning and maintenance.

Summary and key benefits

The steelmaking plant mentioned has reported initial results of 10% less flaring of gases and 15% accuracy improvement of electricity procurement forecasts. Site-wide optimization of energy consumption and availability was a key factor in achieving the results as the solution covered steam yield, by-product gases, energy purchase and production, including site power plants and turbines.

In addition, there are further opportunities to scale and enhance installed digital systems based on the latest edge computing, cloud, visualization, AI and ML technologies; for example, by using analytics for adaptive re-modelling and tuning, including additional variables for optimization.

Taking a wider view, visualizing the top steelmakers’ strategic roadmap versus available technologies ripe for analytics and ML demonstrates just how big the potential is for digitally optimizing energy use in the industry. Policymakers are also looking at best practices to set and track adequate targets.
FIGURE 7: Integrated decision support system to drive continuous performance improvements
A multitude of research is ongoing in areas such as process-specific data models and simulations, carbon capture, utilization and storage, or making green hydrogen a financially feasible option.

Encouragingly, this will help steelmakers shift away from current production methods towards completely new ways of improving energy efficiency, reducing energy cost and carbon footprint.

Digital energy optimization is a perfect use case for both AI and ML, with multiple data sources and concrete problems to solve.

Case study 2: Cement industry emissions

There are other drivers forcing the transformation of process industries. Take new approaches to transport, accommodation and food delivery, as well as changing customer demands. For example, modular prefabricated buildings will use less cement, as well as different types, meaning manufacturers will need to look for efficiencies in the entire end-to-end value chain – from planning to shipping, logistics and production optimization – to retain profitability at reduced cement demand. The industry will embrace new business models around high-tech low-CO2 cements.
FIGURE 8: Cement plants benefit from advanced process control
Here is another way to apply advanced analytics and AI, this time for emission control in a cement plant. As cited earlier, cement production today is responsible for significant proportion of global CO2 emissions, and worldwide demand is expected to grow by 30% by 20405. Action must be taken to avoid this rapid growth becoming a major contributor to climate change.

A cement plant operations team is also constantly worried about deviating from daily sulphur dioxide (SO2) emission limits and associated hydrate consumption, juggling numerous process constraints. Varying properties in the feed and fuel sources – coupled with complex dynamics – make constant manual optimization challenging, however. Manual operators with PID control tend to remain at ‘safe distances’ from process constraints, at the cost of plant profitability. The goal for cement manufacturers worldwide, therefore, is to standardize their plant optimization strategy with the aim of minimizing shift-to-shift variations and human workload.

This is where advanced process control (APC) comes in, handling complex multivariable processes to constantly tweak the production process into an optimum state, and keep it there for as long as possible.

The APC controller reduces operator workload by automatically optimizing the short-term exhaust SO2 target based on the current daily average. The operator can select one of two optimization modes – ‘normal’ and aggressive’ – based on whether they want to target the daily average below the limit at the end of the day or within the next 30 minutes. The APC solution from ABB then automatically adjusts the multiple feeder points of lime hydrate to ensure SO2 and HCL targets are strictly met4.

Looking into a practical application for AI/ML, how do we tie analytics and advanced process control (APC) together in an automated way? What are the benefits of doing this? Let me give context to the problem using a typical phenomenon experienced on site with an existing ‘classic’ APC solution:

Traditionally, APC relies on model predictive control and moving horizon estimation strategies that use either a linear or non-linear mathematical model of the industrial plant and smart algorithms to estimate unmeasured states and control process variables. APC helps industries attain operational and financial targets by increasing throughput and reducing energy use.

Typically, process industries and energy companies integrate APC in distributed control systems, which allows industry users to benefit from distributed resource allocation, redundancy, and communication as well as the intrinsic cybersecurity infrastructure of these modern DCS.

However, as APC technology continues to evolve with new components and features, so does the potential of AI with the use of reinforcement learning neural networks as well as edge and cloud technologies for digital analytics and optimization for operational services in the process and power industries.

With advanced analytics, AI and reinforcement learning from the neural network, processes can be further automated so that the performance and accuracy of the models are continually monitored, and analytics provide new models for the controller. This includes open-loop systems, when the recommendations are provided to a human operator, and closed-loop systems, when the core software rewrites the best setpoints learnt by AI directly into the control system. This will reduce engineering time required to perform these tasks and allow the system to stay at peak performance, maximizing the profitability for cement producers while reducing emissions.
FIGURE 9: The move towards adaptive APC for re-modelling and tuning, optimizing additional variables

Summary and key benefits

In the cement industry example, ABB’s APC software minimized standard deviation, significantly reduced deviation around the operating target for daily SO2 emissions and helped reduce overall hydrate consumption in the plant by 11%. Operator utilization of the emissions controller is 91%6.

This final statistic is indicative of growing confidence in process industries around digital solutions. In 2021, the human still has ultimate power, but at ABB we develop functions so that operators receive the information they need – and they also know what the APC and analytics are doing in the background.
 
FIGURE 9: The move towards adaptive APC for re-modelling and tuning, optimizing additional variables

Investing in human capital

This successful collaboration between humans and machines is fundamental if steel, cement and other process industries are to reap the full benefits of industry 4.0 innovations when it comes to sustainability.

We, as engineers, are good at domain-specific knowledge, intuition, creativity, empathy. Machines, conversely, are good at large data sets, calculations, learning, automation, pattern recognition. The company that can marry the two successfully will enjoy competitive advantage in the digital future.

Sustainability-related actions mean different things to different people depending on their job function, and they experience varying degrees of difficulty when it comes to locating and analyzing data across their organization. An industrial-grade analytics and AI platform helps you get the most from federated data sources – your existing control systems, IoT devices, MES, ERP, CMMS, engineering and other databases.

Contextual Data Fusion tools automate data integration, letting operators identify previously hidden relationships and performance trends, make timely predictions and accurate forecasts. The business value from taking this holistic view is tailored to specific roles and enables cross-enterprise actions.

However, this can only be achieved if you trust people who are building the optimization models for specific use cases: their deep process expertise, the know-how of your industry. Now, their expertise and ingenuity can be captured and shared on a common platform – augmented with edge, cloud and ML technologies for maximum impact. Industrial analytics helps various functions in all industries to collaborate, augmenting human abilities with AI/ML, and focusing on solving concrete problems.

Earlier, I mentioned investment in people and its importance to the digital transition. In this context, when we talk about technology being ‘disruptive’, we don’t just mean in terms of quantifiable, statistical gains around efficiency and sustainability: we also mean culturally within an organization.

When we talk about the digital transformation, we often do so in terms of futuristic technology, but in fact, if it’s done well, it is more a commitment to continuous improvement and perpetual change – in order to stay relevant. Therefore, I prefer the term DigitalOps. Practicing DigitalOps means continuously discovering, modelling, analyzing, combining, extending, and optimizing business processes in a way that allows companies hit their KPIs around operational excellence, sustainability, process and asset performance, safety and productivity of connected workers. It also means looking differently at cyber security – as an enabler and accelerator of DigitalOps, defending your business in a dynamic, strategic way.

Understandably, many people ask: “Is my job at risk, or will it vanish, as a result of digitalization and automation?” This is the wrong question. Certainly, many personnel in the process industries will see the way in which they work fundamentally transformed, but the real question should be: “How can I best apply and adapt my existing skillset to allow me to be relevant and prosper in the brave new world of digital?”

There is a significant shortage of skilled labour in the process industries. The onus is on companies to demonstrate that they are serious about ESG policies and innovation if they wish to attract the next generation of digitally literate talent to whom issues such as climate change are a genuine concern.

The technology transition must be supported by the transformation of processes and people. How well a specific solution solves a use case is often not as important as how it fits within the overall operational business culture. Unfortunately, successful change management is often overlooked7.

Accelerating the use of AI

AI is transforming industry. One way that ABB develops AI capability is through making low-code/ no-code solutions accessible to more people and letting them capture “tribal knowledge” on a common platform. Amongst other activities, the company collaborates with industry consortia, innovation-driven start-ups, conducts the ABB Industrial AI Accelerator program to drive the next level of the industrial revolution via such collaborations.
FIGURE 11: ABB’s approach for data and analytics and AI in process industries
ABB practices an ‘outside-in’ approach, where external ideas are brought into the company to complement ABB’s own innovation activities. Engaging with external partners, such as system integrators, universities, research institutes or start-ups, enables ABB to both identify and capitalize on breakthrough technologies or new business models that help the company find new offerings for its customers.

AI, often dubbed “the next technology frontier” or “the Intelligence Revolution”, is one area where collaboration can make all the difference. AI has the potential to enhance human capabilities in a range of industries. AI is still a growing field, in terms of venture capital investments it is one of the best-funded sectors. Large firms in almost every industry are trying to integrate AI capabilities into their offerings. According to the AI Index Report 2019, globally, investments in AI start-ups have increased at an average annual growth rate of more than 48% since 2010 (2018: $40.4bn)8.

It is in the interest of all the process industries to invest in their digital future, particularly when it comes to decarbonization and carbon neutrality. The new EU carbon border tax, for example, will apply to electricity, iron and steel, aluminium, fertilizers and cement. Against this backdrop, an effective, coherent sustainability strategy is no longer a ‘nice to have’, it is a business imperative.

Summary and conclusion

In this article, I have discussed how manufacturers in the steel and cement industries can automate contextual integration of operational technology, IT and engineering data, and how to best apply analytics across various processes, sites and functions – moving to an entirely new level of strategic predictions and prescriptive actions that contribute to sustainability and other business goals.

As I have explained, with energy cost, efficiency and sustainability gaining top priority globally – particularly for heavy industrial energy consumers – new digital tools are being developed at a much faster pace than ever before, in collaboration with universities and other ecosystem players. 

A substantial amount of research is ongoing in areas such as process-specific data models and simulations, carbon capture, utilization and storage, and making green hydrogen financially feasible. This will help process industries shift away from current production methods towards radically new ways of improving energy efficiency, reducing energy cost and limiting emissions. 

Today we are able to establish a single source of truth system for all data related to regulatory compliance, emission optimization, energy mix planning, consumption and conversion, covering not only electricity, but fuel gases, cryogenics, cogeneration gases, water, waste, and any critical resources for circular economy and related optimization. Combining perspectives from the shop floor and the top floor directly contributes to strengthening this continuous cycle of improvement.
ABB Digital solutions for sustainability
FIGURE 12: Sustainability GRI reporting examples
In conclusion, technological innovations such as AI, ML and advanced data analytics, provided by a specialist technology vendor with proven domain knowledge, operated by a motivated, digitally literate workforce, and backed by meaningful long-term investment have tremendous potential to optimize energy management and emission control in the steel and cement industries, allowing companies to hit environmental targets and contribute to a cleaner, safer, more sustainable world.
ABB Industrial Software

Sources:

1 https://eciu.net/analysis/briefings/international-perspectives/what-is-cop26-who-will-attend-it-and-why-does-it-matter
2 https://ukcop26.org/wp-content/uploads/2021/11/COP26-Presidency-Outcomes-The-Climate-Pact.pdf
3,4,5 ‘Perspectives submission for Digital Chemical Engineering Journal’
6 Sanjit Shewale – ‘Practical ways to apply data analytics and AI for sustainability’
7 ‘The mining workforce of the future’ – Roze Wesby, ABB
8 https://new.abb.com/news/detail/74694/innovation-and-speed-in-industrial-ai
  • Contact us

    Submit your inquiry and we will contact you

    Contact us
Select region / language