With its thousands of reflecting parabolic troughs facing the sun, the site looks like a giant mirror in the middle of the desert. Located in one of the hottest areas of North America, the solar power plant in the Mojave Desert generates 280 megawatts of clean electricity to power roughly 90’000 homes in the region. A similar picture on the other side of the Atlantic Ocean: KaXu Solar One in South Africa uses the power of the sun to generate 100 megawatts of electricity for the population.

Preventing 665’000 tons of CO² emissions per year

The two solar plants belong to the diversified portfolio of Atlantica, a global player with sustainable infrastructure assets all over the world.

Together, the two facilities prevent the release of 665’000 tons of CO² into the atmosphere every year. That’s the same amount of emissions that forests the size of Yosemite National Park can absorb in one year.

Using the power of the sun

Several energy scenario studies consider concentrated solar power (CSP) to be a key sustainable source of energy to meet ambitious climate protection goals.

CSP systems use mirrors or lenses to concentrate the sunlight onto receiver tubes. These tubes contain a heat transfer fluid that is pumped around the whole site. Together with boiler feed pumps, the heat energy is used to create steam, which drives a generator to produce electricity. Pumps therefore play a vital role in ensuring the functionality of the entire plant.

One step ahead of outages

Because thousands of people depend on this plant for power, the operators are keen to run their plants as efficiently as possible and avoid downtimes. For this reason, Atlantica invested in Sulzer’s BLUE BOX advanced data analytics platform to monitor and optimize the performance of its plant.

BLUE BOX uses machine learning to interpret live pump operating data. Based on this data and with the know-how of Sulzer experts that comes with the solution, the system supports plant operators to optimize operation and maintenance of their pumps.

It detects and flags anomalies, estimates the remaining lifetime of the equipment and helps make data-driven decisions for preventive maintenance.

Sulzer’s solutions play a critical role in infrastructure all over the world where downtime means profits lost. Marc Heggemann, Head of Group Commercial Digital Solutions at Sulzer, explains how artificial and human intelligence together optimize the efficiency of assets and predict equipment failures.

Can artificial intelligence and algorithms predict an outage of a plant?

It needs machine learning, but that’s not enough. Why? The amount of data is often limited, and quality is sometimes not good enough, since operators lack a complete history of pump operation and maintenance. Hence, machine learning needs to be combined with physical pump modeling to increase confidence in the results. That’s where Sulzer comes into play: as an OEM, we already have the required pump knowledge in-house. It takes thorough expert understanding, or human intelligence, of how the physical equipment design and operation are represented in a so-called “digital twin”. Building on that, our equipment optimization specialists support customers with the data analysis required to draw the right conclusions.

How does BLUE BOX work in practice?

BLUE BOX is an early detection system flagging anomalies on key performance indicators of pumps. These anomalies are often not uncovered by threshold alert systems on individual sensors until it’s too late. Contrary to an instant alert, anomaly detection occurs early enough to allow preventive maintenance action. Let me give you a real-life example. After implementing BLUE BOX at Atlantica’s plants, the system flagged four anomalies on a single pump over a couple of days, indicating that the performance of the asset deviated from its healthy state.

This finding was confirmed by Atlantica’s on-site data science team which also found an abnormal event by analyzing the data. The analysis of the motor power and shaft speed afterwards identified a bearing that was close to failure, yet far below the alarm limits for vibration and temperature in conventional methods. The customer was able to proactively order spare parts, mitigating the risk of failure and saving money.

Why should a cost-conscious plant operator invest in such an artificial intelligence solution?

If you look at the total life cycle cost of the equipment, the investment in such AI solutions in relation to potential savings is rather small and makes a valid business case. We add value by supporting our customers in their decision-making processes through customized cost-benefit analyses. The unexpected failure of critical pumping systems can easily have implications exceeding USD 100’000 per occurrence. BLUE BOX monitors equipment and flags anomalies before failure happens, thus avoiding costly downtime and reducing operational risks. Combining these results with our expert knowledge, we can recommend the best and most cost-efficient solution. This allows the customer to extend equipment lifetimes and reduce life cycle costs, typically achieving payback within months.