AI/ML solutions vital for efficient aeroderivative turbine power required to fill renewable variability

As more and more renewable energy sources come online, the need for speed and agility to produce power grows. Changes in weather happen quickly. A sudden decrease in wind or sun can mean traditional power generation must respond just as fast.

This is where aeroderivative turbines bridge the gap to their more powerful counterparts. For example, an aeroderivative turbine can go from start up to full power in as little as 5 minutes. For comparison, a typical medium-speed reciprocating engine can have a ramp rate of about 5 MW/min, whereas the aeroderivative is around 50 MW/min.

How’s that possible? Aeroderivative turbines are essentially grounded jet engines that have been reconfigured to run on natural gas. They can be mounted on a trailer and quickly employed and connected to the grid. This makes them an optimal choice for on-demand power required when renewable generation varies.

And their popularity continues to grow. According to a study by Global Market Insights, the global aero-derivative market is set to exceed $3 billion by 2026, up from $2.3 billion in 2019.

Power agility, but efficiency is elusive

The operability window of any turbine is bounded by emissions, lean blowout, acoustics, fuel variations and drift in fuel valve calibration. Gas turbines tuned on a particular day may go out of their operability window as ambient conditions and seasons change. This can require multiple seasonal tunings to bring it back into emissions compliance.

Additionally, fuel variations can impact machine performance relative to its operability boundaries. Finally, load and mode combination vary with ambient conditions and fuel composition and also with the control settings provided by the tuning engineer. 

AI/ML for maximum efficiency

Just as aeroderivative turbines serve a vital role in the energy transition, software is critical to ensuring those turbines run as efficiently as possible.

By sensing changes in ambient temperature, gas fuel properties and degradation, a program can send real-time adjustments to the turbine controls. By applying as Level 2 software, AI/ML can be fully bound by the controls system safety-critical programming to ensure no harm to the turbine.

GE Digital offers such an AI/ML solution. Autonomous Tuning is an on-premises solution that sends adjustments to the controls every two seconds. It’s part of a larger platform strategy of Advanced Combustion Control. By moving from static, point-solution maps for the combustion system to a dynamic, adaptive, and model-based system, power generators enjoy the benefits of automated tuning and have the foundation to upgrade the flexibility of their unit with emissions-compliant base load operation, emissions-compliant turndown and overall part-load efficiency.

For power generators, GE Digital’s Autonomous Tuning solution addresses a number of challenges:

How Autonomous Tuning works

Autonomous Tuning has two automated modes of operation – Learning Mode and Control Mode – that are linked by a human-supervised model building step.  

The modes of operation are sequential. Learning Mode must be executed first to map the space of operation of the turbine. Data collected in Learning Mode is then used to build a neural network model of the turbine’s behavior. Once the models have passed quality checks, they are used in closed loop to adjust the turbine’s flame temperatures to ensure optimal behavior. The goal is to allow for tracking of the turbine’s sweet spot (operational conditions with low acoustics and low emissions) in response to changes in environmental conditions, fuel properties, or physical degradation, and reduce the need for seasonal retuning. 

Benefits

Customers are already reaping hard savings from Autonomous Tuning, including:

• 0.5% to 1% reduction in fuel consumption / CO2 emissions
• Up to 14% reduction in CO emissions
• Up to 12% reduction in NOx emissions
• 0 manual tunings or associated downtime

AI/ML is not optional for Energy Transition

For energy producers to achieve decarbonization, every tool in the toolbox must be applied. Aeroderivative turbines are a part of the solution to bringing more renewables online while satisfying the insatiable need for power. But without digital solutions to ensure they’re continually optimized for lower emissions and fuel consumption, they cannot fully contribute to the energy transition. Digital solutions are no longer optional. The Energy Transition demands we employ every measure for efficiency. This is good news for power generators. Lower emissions and fuel consumption from AI/ML optimizes O&M spend as well.