By Don Clayton, Supervising Engineer, AmerenUE, and Rob James, Product Manager, NeuCo Inc.
A fter four months of effort, nothing was working,” said Tom Ziegler, AmerenUE performance engineer, referring to the time 13 years ago when AmerenUE first tried neural network-based combustion optimization technology at its Labadie plant to help reduce its nitrogen oxide (NOX) emissions. “But instead of cutting our losses, we kept trying.” Two years later, Labadie had one of the lowest NOX emissions rates in the nation among coal-fired plants with no selective catalytric reduction (SCR) systems. The then-novel combination of combustion optimization software, low-NOX burners (LNB), over-fire air (OFA) and in-house expertise was the key reason.
AmerenUE was also one of the first utilities to install a commercial OFA system on a cyclone unit, which was previously deemed infeasible. For this, the company received the 1998 Missouri Governor’s Pollution Prevention Award for achieving unprecedented NOX reductions at its Sioux station. These two examples represent the prevailing culture at Ameren: one of environmental stewardship, innovation and continuous improvement.
AmerenUE (previously known as Union Electric) operates four large coal-fired power plants near St. Louis, MO: Labadie, Meramec, Rush Island and Sioux. Over the past few decades, AmerenUE has taken steps to meet the challenging regulatory and business environment faced by the power generation industry. In addition to expanding the use of combustion optimization, low NOX burners and over-fire air to AmerenUE’s full fleet of coal-fired plants,
AmerenUE was among the first Midwestern utilities to switch to low-sulfur western coal beginning in the early 1990’s. Starting at Labadie, the switch to low-sulfur Powder River Basin (PRB) coal was ultimately implemented at its other three plants and AmerenUE became among the largest users of PRB coal in the nation. Figure 1 shows a timeline of the company’s emissions reduction efforts.
The results of these efforts are illustrated by the large reduction in sulfur dioxide (SO2) and NOX emissions since 1990 (Figures 2A and 2B). The company’s combustion optimization efforts, which initially used the Pegasus Technologies’ system, were particularly successful at the Labadie and Rush Island Plants. For the last decade these plants have ranked in the EPA’s top 20 coal-fired power plants in the nation in NOX performance for units without SCR.
Meramec Plant Focus
In 2007, AmerenUE decided to pursue further emissions reductions and performance improvements at Meramec. The four-unit 930 MW plant, which began operation in 1953, is on the Mississippi River near St. Louis. Units 1 and 2 have tangential-fired Combustion Engineering (Alstom) boilers originally rated at 135 MW. Unit 3, rated at 290 MW gross, has a front wall-fired Foster Wheeler twin furnace boiler. Unit 4 has a front wall-fired Foster Wheeler boiler rated at 370 MW gross.
AmerenUE’s success with combustion optimization at Meramec Plant was pronounced at Units 1 and 2, but not Unit 4 and had not been attempted at Unit 3. Combustion controls at Unit 3 were problematic because the unit did not then have a distributed control system (DCS). Its boiler design made using this technology
challenging because it included two separate furnaces feeding one set of emissions monitors, with NOX , CO and other instrumentation relating to the complete unit instead of each furnace.
Beyond NOX to Improved Operations
NOX reduction was still a top company goal, but Meramec also had specific operational needs that went beyond NOX : to keep the units running reliably and to push them to get the best possible megawatts and maximum efficiency by maintaining optimal boiler cleanliness.
By this time optimization software provider NeuCo Inc. had acquired Pegasus Technologies. In addition to providing a newer combustion optimization system, NeuCo provided a total boiler optimization package, which also included sootblowing optimization.
Sootblowing optimization was important to Meramec because the plant had a lot of erosion-related tube ruptures, especially on Unit 3. These ruptures caused forced outages. Unit 3 also had significant temperature issues: air pre-heater (APH) gas inlet temperatures were very high at full load and constrained capacity. On Unit 4, carbon monoxide (CO) and a significant imbalance in reheat steam temperatures between the east and west sides were the chief concerns. If such a split is too high, it can cause thermal damage to the turbine. On Units 1 and 2, reheat and superheat steam temperatures tended to drop too low at low loads. This negatively affected NOX emission control.
Getting Started – Unit 3 Implementation
Meramec management decided to implement the complete NeuCo boiler optimization package (CombustionOpt and SootOpt) on all units, starting with Unit 3. The CombustionOpt project kicked off in February 2007. The initial goal was to reduce NOX emissions; however, a number of boiler constraints made that difficult, particularly the need to maintain stack CO at acceptable levels. There was also no control for windbox pressure.
CombustionOpt uses neural networks, model predictive control and other technologies to extract knowledge about the combustion process and determine the optimal balance of fuel and air flows in a furnace. It optimizes fuel-to-air ratios in real-time by biasing DCS setpoints to adjust dampers, burner tilts, pulverizer settings, over-fire air and other controllable parameters to their optimal levels for a given set of conditions, objectives and constraints.
Unlike typical NeuCo optimization projects, Ameren had an in-house resource, Tom Ziegler, who was expert at implementing neural network combustion optimization systems. Since this current project introduced a new software platform, NeuCo engineers implemented CombustionOpt on Unit 3, while training Ziegler and a second Ameren engineer, Scott Hixson, to take the lead on the other CombustionOpt projects.
On Unit 3, 53 controllable parameters were configured for bias control including excess O2, SOFA damper, shroud damper, Mill PA flow and FD duct pressure biases. The DCS logic tie-ins and the software configuration were completed in June 2007 and the system was placed in closed-loop neural network control. Further modifications were made by NeuCo and Ameren engineers and the project transitioned into customer support.
The primary goals of adding sootblowing optimization on Meramec Unit 3 were to help operators maintain superheat and reheat steam temperatures within a narrower range, reduce air pre-heater gas inlet temperature excursions on the superheat furnace and reduce sootblower long-lance operations. Meramec plant management anticipated that these goals would improve heat rate and promote tube longevity.
SootOpt is a closed-loop sootblowing optimization system that uses expert systems to model the effect of sootblowing activity on heat transfer throughout the furnace and backpass. It dynamically determines the boiler cleaning actions required to improve temperatures, heat rate and availability and minimize NOX . At Unit 3 it works in conjunction with an existing programmable logic-based, sootblowing control system.
SootOpt uses a two-step decision making process to determine which blower, if any, should run. The first step is to select a zone, which consists of groups of sootblowers logically related by their association with a particular part of the boiler. Zones are selected through rules, which are in turn based on the principles of the heat exchange processes within the unit. The decision is also influenced by each zone’s recent sootblowing history.
The second step is to decide among blowers within the selected zone. SootOpt can either choose the eligible blower with the greatest idle time or choose the one, based on neural network predictions, that indicates potential for the greatest improvement in unit operation.
NeuCo worked with Meramec engineering and operations personnel to develop the unit-specific expert rules that describe Unit 3’s heat transfer characteristics. This expert knowledge was embedded into the initial rules configuration. This “knowledge base” can be changed and expanded over time by modifying rules to reflect the latest available instrumentation and wisdom.
The initial CombustionOpt “on-off” test showed minimal NOX reduction, but results improved as the system was tuned. Meramec Unit 3’s average annual NOX rate in 2007, before optimization, was 0.196 lbs/mmBtu; its average annual NOX rate in 2009, after CombustionOpt had been running for a year and SootOpt had been running for 9 months, was 0.177 lbs/mmBtu, a reduction of about 10 percent.
Minimizing Tube Erosion
SootOpt helped to reduce and balance Unit 3’s sootblowing activity, especially on the long lances that were known culprits for their contribution to erosion issues and associated tube leaks. Before, operators were manually blowing three lances 40 to 45 times in every 24-hour period; this was cut almost in half with SootOpt. Overall, the reheat section long lance operations were reduced by about 33 percent and the superheat long lances by about 7 percent.
Another benefit was the fact that operators trusted the system. They could spend less time managing sootblowing and monitoring exit gas and steam temperatures and more time on other priorities.
SootOpt was not the only investment Meramec made to reduce tube ruptures. The plant also performed boiler inspections to identify the areas with the most erosion and invested in tube shielding and pad welding in the worst areas. The knowledge gained through this process was captured in SootOpt’s rules and constraints.
Improved Temperatures & Restored Capacity
The optimization system also helped reduce exit gas temperature excursions over Unit 3’s high limit of 860 degrees, which, in turn, helped to increase its maximum output by up to 25 MW. The superheat and reheat steam temperature ranges were also narrowed, improving temperatures while reducing attemperation sprays.
For Units 1, 2 and 4, NeuCo completed the SootOpt installations while Ameren led the conversions from the Pegasus combustion optimizers to CombustionOpt. All systems are now running in closed-loop control.
The on vs. off NOX analyses for SootOpt on Units 1, 2 and 4 were performed without combustion optimizers enabled. Unit 1 NOX across load profiles was reduced by an average of 14 percent, with a range from an 18.5 percent reduction to a 5.6 percent increase. On Unit 2, NOX stayed about the same at high loads but was reduced by up to 40 percent at low loads. On Unit 4, there was a modest NOX improvement at mid and high loads.
For Units 1 and 2, sootblower actuation counts for the long lances and wall blowers were reduced by similar average percentages: 38 percent and 26 percent, respectively. However, the goal of sootblowing optimization is not always to reduce counts. On Unit 4, for example, the sootblower actuation counts for the APH long lances had to be increased to obtain optimal heat absorption in the air heater.
On Unit 4, SootOpt helped to minimize the split in reheat steam temperatures between the east and west sides, from 25 degrees to 8 degrees. Air pre-heater gas inlet temperatures were also improved with 18 and 5 degree reductions on the A and B sides respectively. On Units 1 and 2, steam temperature improvements focused on the reheat section since low reheat steam temperatures at low load had been one of the key challenges for these units (Figure 3).
With the dust not yet settled on Meramec’s new boiler optimization strategy, the recent changes are not the last leg of AmerenUE’s optimization journey. Ziegler and the other AmerenUE engineers continue to find new ways to improve operations. At Meramec, they are currently testing the model predictive control (MPC) component of CombustionOpt, which is a predictive controller that allows for faster cycle times. They hope to improve O2 variance during load changes and better control CO while ramping up.
Meanwhile, Sioux is testing the full suite of NeuCo optimization systems on one unit, combining boiler optimization with performance and equipment anomoly detection and diagnostic sytems.
From being an early adopter of environmental control and neural network-based combustion optimization to implementing and evolving a total boiler optimization solution, AmerenUE continues to push the envelope.