Engie has been developing Darwin, its own remote monitoring and predictive maintenance application, for the past seven years. In 2010, the French energy company opened its Renewable Energy Management Centre (REMC) at Châlons-en-Champagne in eastern France, which would become Darwin’s base.

“We had a wind farm in France that in 2010 we decided to connect to the distribution grid, so we needed 24/7 remote surveillance,” says Damien Terrié, head of the Darwin Project.

“We started to develop what became the Darwin software to do that.”

The inspiration for Darwin was to provide a standardised way to remotely monitor and analyse equipment that was independent of each manufacturer’s often proprietary supervisory control and data acquisition (SCADA) systems. With its ability to connect to virtually any automation system or device, Schneider Electric’s Wonderware SCADA suite was ideal for this.

The system is the backbone of Darwin, but the vast majority of subsequent development work has been run in-house. Working in conjunction with specialist subcontractors, the development team is now 40 strong.

Each farm’s SCADA systems connect to an on-site server, which in turn connects to Darwin’s cloud-based platform. Most modern turbines come equipped and ready to plug in but legacy fleets acquired from other operators can require some retrofitting.

“Typically, we try to install the least possible equipment for data acquisition but sometimes we have to,” says Terrié. “For old turbines, we do have to add additional monitoring equipment, particularly for vibration. Because these wind farms are globally based, it’s become a bit of a supply chain nightmare.”

Measurements like wind speed and active power are sampled at one-second intervals and transmitted in real time. Other metrics vary far less and are typically collected and processed on-site before sending. For example, gearbox temperature would be sent at ten-minute intervals.

Creating value through data

Engie is also active in adding internet of things (IoT)-enabled sensors to capture more information on local weather conditions. One application is logging air density directly at the turbine nacelle rather than taking a measurement from a nearby weather station. Rain sensors are another common addition.

The inspiration for Darwin was to provide a standardised way to remotely monitor and analyse equipment that was independent of each manufacturer’s SCADA systems.

“We keep innovating with what we can put on top of the turbine to make the wind farm smarter,” says Terrié. “Air density is directly linked to the power available in the wind so it lets you check your power curve more accurately.”

Data typically starts its journey from the farm via a local broadband connection. However, other means are sometimes necessary.

“There’s not that much you need to send in true real time, so we don’t need gigabytes per second,” notes Terrié. “We try to avoid satellite connections because they are so expensive.”

To the field, data pouring in from tens of thousands of global feeds is added other external information such as detailed local weather forecasts and energy market pricing.

Maintenance teams use a tablet app to input their reports directly into the Darwin database.

All that data is processed and stored using the MS Azure cloud platform. Managing the terabytes generated is challenging – as is making full and timely use of it to improve operational performance.

“You really have to bring value to the people in the field,” says Terrié. “There’s no point in telling them something they know already. Darwin can detect an overheated gearbox, whether turbine blades are covered with a layer of ice or whether it is necessary to clean solar panels.”

To supply local operation and maintenance (O&M) teams with information, the company employs a browser-based interface customised to each region’s needs. Using responsive design means it will work across multiple devices irrespective of screen size. When the team members come into the office in the morning, they can look at personalised aggregated reports and see how the turbines performed overnight or today’s wind speed forecast.

When problems do occur, Darwin alerts them to it, using its automated insight to suggest the appropriate corrective action, and then helps choose when to perform it. One example might be identifying a fault that needs to be fixed within the next 30 days. The app helps the operations manager to pick a time when his maintenance team is available, and the wind speed and the spot market electricity price are low – so minimising the impact any downtime has on profit.

Regional data labs

Data scientists are the ones who build automated analyses like predictive maintenance and underperformance detection in Darwin, using sophisticated predictive modelling and other analytic techniques. Each region has a dedicated “data lab” that gives them access to the full historical database.

From here, they can consider the entire lifespan of a wind turbine, and compare different turbines and farms. Which ones generate the most power from the same wind speed? And which have the lowest overall cost of generation? Terrié points to yaw control as a common example.

“By looking at the nacelle alignment together with the wind direction, we are able to detect any misalignment and see how quickly the nacelle gets aligned into the wind,” he says.

Any problem with yaw control inertia might simply be an individual control system issue that needs to be fixed. But it might also be endemic to a particular model or manufacturer. Darwin has identified wind turbine settings that caused a delay in orientation response, leading to a 0.5% loss in production. Over ten years of operation, correcting that error would be worth €10 million. Similar calculations apply to everything from pitch control to gearbox reliability.

“We know which manufacturer performs better than another or which turbine we should buy – or stop buying,” says Terrié.

Comparing the actual behaviour with theoretical and average behaviour based on the fleet’s history is a vital part of analysts’ work, letting them detect deviation from the benchmark or from how certain turbines – or solar cells – have behaved in the past.

“We look at the power curve of the wind turbine and check that it behaves as it should,” explains Terrié. “Does it cut in where it should, does it reach the nominal power? That sounds easy to do but with the amount of uncertainty in the nacelle anemometry, it’s quite tricky to be able to generate a clear signal. In the turbines where we manage to do that, it really brings value.”

Predicting output and performance

Darwin’s intelligence supports non- Engie users too, with an application programming interface (API) that exposes data feeds as web services so that external web apps or other user applications can consume it. Exporting production data to participate in energy markets is one vital application here.

To the field data pouring in from tens of thousands of global feeds is added other external information such as detailed local weather forecasts and energy market pricing.

By combining that information with weather forecasts and spot market pricing data, it’s possible to feed information to external or Engie’s own internal market aggregators in order to predict output and so bid optimally when they commit to a volume of production.

“If you know exactly how much you are generating now, you have a much better idea of what you will generate in the next 15 minutes,” explains Terrié. “By accessing the data in real time, they can lower their intra-day market risk and so the premium we pay is smaller.”

Via a website, smartphone apps and a chatbot, it’s possible for the public to see real-time information on generating facilities and their performance. That might be a factory owner whose site has Engie’s wind turbines in the car park or solar panels on the roof, local residents near the wind farm or local authorities.

“The factory owner can know in real time how much he is contributing to the energy revolution and he can display that on his own website,” says Terrié.

Darwin’s business proposition contains the hard figures. A system investment of €13 million over two years has been balanced by extra income of €27 million over three years. Knowing how valuable that information is to the business, Engie is rapidly connecting the rest of its renewable assets to Darwin. Around two thirds have been linked so far, with the rest due by the end of 2018.

“We’re going after the big fish first, connecting our largest assets then the smaller ones,” says Terrié. “The readiness of the local team to use the Darwin information also has an influence.”

The development work continues, with a February 2017 agreement between Engie and Schneider Electric to further develop remote supervision and control of its global renewable energy assets. Asset management, SCADA obsolescence management, remote monitoring, diagnostics and cybersecurity will all be investigated.

Further showing its confidence in the value of the data it generates, Engie recently made all the historical operational data from its small La Haute Borne wind farm fully public. Using this open data will showcase Darwin’s capabilities to developers, manufacturers, operators, researchers and students.

“Five years ago, it was very difficult to look at your data like this,” says Terrié. “Darwin really is a revolution.”

What is an OPC server?

Built by vendors like Matrikon and Cogent, open platform communications (OPC) servers act as central hubs on wind or solar farms. They connect securely to the SCADA modules linked to the control and automation systems as well as other equipment such as the power management unit or the weather station.

OPC servers gather this dispersed data centrally and then make it available in real time for remote control and performance monitoring. OPC servers support connectivity with a wide range of external systems, from handheld tablets and data loggers to cloud platforms like Darwin.

These servers are also known as IEC servers. IEC 61850 and other allied standards provide open protocols that support monitoring and control of wind power plants along with many other electronic devices. The interoperability this standard confers allows the farm owner to exchange information with different wind power plants independently of a vendor’s proprietary systems.

Collaborating on turbine reliability with SPARTA

Formed in 2013, the SPARTA joint industry project aims to help improve the availability, reliability and performance of the UK’s offshore wind assets. Operational data is collected using agreed standard definitions at system level, from blade to onshore substation, with each participant submitting monthly metrics to a secure server for each of its wind farms. These are aggregated at project, portfolio and sector level, and shared between the eight contributing owner/operators. With all the market leaders involved, the high-level data so far released for the year to May 2016 covers a wide range of farm ages and locations as well as turbine types (3.55GW, 1,045 wind turbines).

Learnings include that younger, larger farms further from shore are currently achieving lower availabilities than older projects closer to shore. Other metrics suggest that as the wind farms mature, increased maintenance experience and improved work processes lead to higher availability.

However, far-shore sites’ capacity factors are higher, which is generally attributed to increased wind speed at those locations and more modern turbines. As these assets mature and availability increases, higher production levels should result. The average monthly repair rate is 1.32 repairs per turbine per month, with the top three subcomponents requiring intervention being hydraulics, the blade adjustment system and rotor.