Software CEM operates in real time using existing process sensor data. These process values enable the plant to monitor operating conditions that could affect final emissions output. Software CEM uses a patented sensor validation system as a qualifier to detect sensor failures and set appropriate alarms. The system uses existing sensors to generate a model of all sensors in the process. This allows data validation to continue accurate emissions predictions during a sensor failure, providing for near 100% uptime. This predictive methodology also gives AECC the ability to simultaneously incorporate process behavior and feedback into the control strategy of its gas turbines.
"After we trained the model," Bivens says, "we're consistently getting data that replicates that from the hardware-based CEMS. We still have it installed and are still reporting its data while we wait for certification for the software CEMS."
Miller continues, "The use of hybrid modeling through empirical models and first principles knowledge gives AECC an excellent representation of its process behavior." Software CEM also has the versatility to predict emissions in the extreme operating ranges of unit operations, he adds.
When AECC initially considered a hardware CEMS, it understood the challenges of ultra-low NOx emission limits that exist with high signal-to-noise levels on analyzers. The result is often poor readings and potential NOx absorption into the sample line.
A heated sample line can have as much as 1-ppm to 2-ppm NOx absorption. On 40-ppm low-NOx applications, there is no effect. However, for ultra-low-NOx applications, where the NOx is less than 5 ppm or 10ppm, it becomes a significant issue for hardware-based CEMS. With Software CEM, these problems are alleviated, since there are no analyzers or system samplers required to predict emissions in extreme operating environments.
"The actual install for the Software CMS cost more than installing new hardware-based monitors would have, but over a five-year baseline, it will be considerably cheaper," Bivens says. "Plus, we free up an instrument tech that had to spend all his time working with the hardware, and there's no quarterly maintenance necessary. We estimate the payback time to be between five and seven years."
The Oswald generating facility's seven combustion turbines each have 15 points where sensors monitor the process for variables such as fuel flow and temperature and pressure. The existing CEMS consisted of 21 hardware analyzers and Teledyne Monitor Labs' RegPerfect software for reporting.
"There was sort of a missed communication. We didn't know that Pavilion had a reporting package, so we decided to stay with RegPerfect," Bivens says. "Rockwell did the necessary software interconnection, so that the data from the software CEMS goes into the RegPerfect database just as well as the hardware data does."
"Because Pavilion was a stand-alone company before being acquired by Rockwell, we can interface with many different kinds of software, not just our own reporting package," says Miller.
Passing the Audits
"This was our first foray into meeting 40CFR Part 75 with the Software CEMS," Miller reports, "and we needed to make sure that we met an accuracy level of below 7.5% when compared to the hardware CEMS. The EPA requires 720 hours of side-by-side testing to validate a new system for CEMS."
"This was really hard to do because Oswald is a peaking plant," Bivens notes. "The plant comes on in the morning, stays on for peaking and shuts off in the afternoon and goes back on in the evenings. We run only 10 to 14 hours a day."
This made it hard to get that 720 hours. "Especially," Bivens says, "because the hardware CEMS would go down and need to be repaired. Then we lost our side-by-side time."
Miller add, "So we used the historical data from the Historian software package to get the 720 hours, and the EPA agreed to accept this. This way we'll never have gaps in the comparison data."
Bivens adds, "When the hardware works, we have performed a relative accuracy test audit (RATA) twice now, and both times certified at better than 7.5% relative accuracy, which surpassed the U.S EPA, CAMD requirements of ±10%. The initial RATA indicated that the two systems were within 7.5% of each other. However, our most current data shows that they are now within 5% or less of each other. That's well below the EPA requirement.
"We're continuing to use the hardware CEMS for reporting, because we're still waiting for the EPA to send us the certification for the software CEMS," Bivens says. "For the first year, we still have to do a RATA comparing both systems. We did the first RATA in August of 2010, and were able to certify both systems to the EPA specifications. We did a second one in July of 2011, and certified both systems again. We are waiting for the EPA to permit us to retire the hardware monitors, and use the software CEMS only. We expect to hear back from them any time now."
In addition to helping prevent fines for non-compliance, Miller reports that Software CEM will help AECC save more than $50,000 per year in costs relative to operating a hardware-based CEM system. Further cost savings will be achieved over the life cycle of the project due to the higher reliability and lower maintenance costs of the system.
"We haven't seen any performance benefit yet," Bivens says, "because we're still running the hardware monitors and CEMS. When the hardware comes out, we will see what the data can tell us. We're looking into it.
"Unfortunately, out of our five facilities, only two are feasible for the software CEMS because the others do not have enough instrumentation on the process to be able to accurately feed the model.
"But, we're considering expansion to one or two of the other plants, pending approval by the EPA," Bivens says.