On-line optimization: Cajun captures boiler efficiency benefits

Issue 10 and Volume 78.

Peter Spinney, NeuCo

Like most power plants today, NRG Energy’s revenue and profitability depend directly on the ability to sell power on the wholesale electricity market and the costs of production for each unit. Since the value of Cajun’s generating assets is strongly affected by competitive position, it has placed a great deal of emphasis in recent years on efforts to improve (reduce) heat rate.

ProcessLink is installed on Units 1 and 2 at NRG Louisiana Generating Big Cajun II, which are two 575 MW pulverized coal-fired units. Units 1 and 2 operate in a mix of base load and intermediate modes, with frequently changing loads and a historical capacity factor of approximately 60 percent. Boiler efficiency increased by more than 0.5 percent and NOx emissions dropped an average of 15 percent at both units.

NeuCo’s ProcessLink is an on-line optimization system, combining advanced direct-search and neural network-based technology to optimize performance at fossil-fired generating stations. It optimizes NOx and and boiler efficiency in real-time, subject to CO and other operating constraints as determined by each client. ProcessLink also includes a technology platform for optimizing other plant processes through modules that can be added later.

The Big Cajun II units had historically exhibited high variability in excess O2 levels as measured by the oxygen probes across the back of the furnace. Historical efforts to balance fuel and air distribution had only limited success. On-line optimization improved control and reduced O2 variability, providing opera-tors with the means to comfortably reduce excess air while maintaining required safety margins.

Models for boiler efficiency, CO, O2 and NOx were initially built using the data collected during parametric testing. Once installed, however, the models continued to learn and improve over time. The models are retrained daily, combining results from the previous day with the initial training data. This ensures that the models remain accurate, even as the plant’s response drifts over time.

Operators can select the goals for the optimizer (e.g., maximize efficiency or reduce emissions) and can also choose which operating parameters can be adjusted within which constraints to achieve the improved performance.

The data generated as the units are repeatedly optimized is used by ProcessLink to continually improve the accuracy of its models, allowing tighter control. For example, ProcessLink identified that the accuracy and variability of O2 measurements were directly impacted by the reheat and superheat (RH/SH) proportional damper positions. Since these inaccurate measurements were being used to drive the control system, extra leeway had always been provided in the operating region to account for this variability. ProcessLink was able to compensate for this error in the O2 measurement and so the O2 setpoint could be lowered safely, resulting in average boiler efficiency increases and NOx reductions.

Fuel savings from the boiler efficiency benefits alone have provided Cajun Electric with an approximate payback of one year, and these benefits are increasing over time as experience is gained with the technology. Using a conservative $1,000 per ton to value NOx reduction, the combined fuel savings and emissions reduction benefits achieved exceed $1 million per year for each unit.

For more information, visit