Neel Parikh, Key Expert - Performance Optimization, Siemens Energy, and Tara Singleton, Utility Engineer, Jeffrey Energy Center, discussed the set-up, implementation approaches, and the use of closed-loop optimizer solutions.
Folks from Siemens Energy and Jeffrey Energy Center, a smart power plant in Emmet, LA, hosted the session, Make the Most out of Artificial Intelligence (AI) and Machine Learning (ML), at POWERGEN 2024. Neel Parikh, Key Expert - Performance Optimization, Siemens Energy, and Tara Singleton, Utility Engineer, Jeffrey Energy Center, discussed the set-up, implementation approaches, and the use of closed-loop optimizer solutions.
Jeffrey Energy Center
The Jeffrey Energy Center is a power base and coal-fired power plant with three units (800 MW) from the late 70s/early 80s. Optimization applications were implemented in two of the three units.
Siemens Energy’s Optimization Solutions
Siemens Energy’s suite of optimizers is available for many different applications, but specific to Jeffrey’s, the plant was outfitted with a combustion optimizer, soot-blowing optimizer, and temperature optimizer.
The Setup
No mechanical changes were made to the units. “We looked at how to optimize the system in a closed-loop control mode on an ongoing basis, [models will continually update themselves],” Parikh said. The setup included an optimizer application server and Thin Client for each unit.
Singleton said the installation process was simple and from start to finish the hard install was not complex.
Each unit can offer different criteria based on the operating profile. The optimizer has an online learning capability such that it's able to take into account process conditions, wear and tear, changing fuel conditions, etc.
Once the model was established, the team also went through a testing phase. “Parametric testing refers to where we would start making changes via the optimizer and seeing how things behave, how they reacted, and fine-tuning that model based on what we saw,” Singleton said
Validation and Results
The team at Jeffrey’s had several goals and targets it wanted to reach with the optimizers, i.e., “The Combustion Optimization Opportunity.” This includes increased efficiency and a reduction in NOx and CO2 emissions.
The Omnivise Combustion Optimizer targets optimum zone operation while considering dynamic operation constraints.
The combustion optimization approach aims to:
The optimizer uses historical and real-time data, so it’s constantly updating. “It’s a living, breathing thing,” Singleton said. This also includes incorporating all of the data points beforehand (equipment specifics, etc.) Further, “The optimizer also helps manage a lot of priorities all at once,” Singleton said.
Omnivise ML Modeling
ML models adapt over time to changing operating conditions and fuel. The iterative optimizer creates what-if scenarios and identifies the best settings.
“The combination of two, the AI/ML model along with the iterative optimizer is what helps us accomplish this,” Parikh said.
As a note, the optimizer does not replace existing controls (plant DCS), as the optimizer is an addition/modifier, not a replacement.
Optimizer Benefits
The optimizer reduced:
It also enabled more symmetrical reheat steam temperatures and lowered soot-blower air usage. Additionally, all furnace differential pressures remained good.