Power Control: Advanced Process Control
Control and stabilization of a multiple boiler plant can be improved by using an advanced process control (APC) approach.
Coal-fired boilers are used in many industrial processes, the most prominent being steam generation for utilities and power plant turbines. The primary objective of a boiler is to achieve optimum operating efficiency with high reliability and low cost. Techniques to improve boiler operation in the past decade have largely focused on optimizing boiler performance while simultaneously adhering to environmental constraints. Model predictive control (MPC) has become an increasingly popular replacement for traditional boiler control and optimization.
Advanced process control (APC) and MPC have been used in industry to stabilize and optimize plants. Stable control on boiler key combustion parameters balances the burner management system, reducing thermal stress factors and equipment degradation from unstable control.
Environmental legislation requires power generation utilities to reduce harmful gases such as nitrous oxides (NOx) and carbon monoxide (CO) emissions from coal-fired boilers. Tight dynamic control of variables is an interactive process that requires a multivariable controller; this process can increase boiler efficiency and reduce production of NOx. Model-based APC was implemented in a coal fire power generation plant to adhere to regulations while maintaining the required load and efficiency. The results were a 48% reduction in NOx emissions and a 75% reduction in CO emissions.
Multiple boiler configuration plants can offer opportunities to improve efficiency. Automatic controllers with cost-objective functions for boilers can shift loads accordingly to maximize efficiency and reduce energy costs.
In this case study, a chemical company needed to find the most efficient operating point for each of its two utility boilers that were generating steam. Analysis determined the boiler fuel to steam ratio and the marginal production curve. The boiler fuel to steam ratio was found to be quasi-concave. As described below, based on these relationships, the company used APC and optimal load switching on these boilers and decreased fuel cost by 3%.
The MPC control scheme reduced the variability of the steam temperature three to five times better than the proportional-integral (PI) controllers. Regulatory PI controllers provide adequate steady state performance; however, APC can handle interaction and provide more responsiveness and stable operation.
The chemical company’s boiler house consists of seven 20 ton/hr fire tube boilers feeding into a common steam header. The primary objective of the boilers is to produce low- and high-pressure steam for utility usage, which includes continuous and batch processes. Production requirements of various downstream processes, particularly batch processes, cause large changes in steam demand. The boilers use lignite coal that is stored in a coal bunker adjacent to the boiler house, and the calorific value (HHV) of coal is calculated.
The coal is transported from the bunker to storage bins located at the top of each boiler by screw feeders. The coal then gravitates into two stoker hoppers situated on the left and right side at the front end of each boiler. Each stoker hopper feeds coal onto a grate, which conveys the coal through the boiler, forming a coal bed. Coal is ignited by ignition arches as it moves from stoker to the grate. The length of the coal bed on the grate is determined by the speed of each stoker. The use of proportional-integral-derivative (PID) controllers on the stoker speeds yielded little success in achieving performance goals. Each boiler has two forced draft (FD) fans situated on the left and right side at the rear end of the boiler. The FD dampers control the forced draft into the boiler and are used to regulate the furnace pressure. The flue of each boiler is equipped with an oxygen analyzer and induced draft (ID) fan that controls the boiler steam pressure. To prevent flames from burning outside the boiler, the boiler must be kept at a slight negative interior pressure. The ID damper essentially determines the steam demand of the boiler.
The objective of the base layer control is to stabilize the process using PI feedback control loops. The FD dampers control the forced draft via pressure PI controllers. Sticktion (mainly from the coal ash in the boiler house) has a tendency to accumulate on the final control elements of the FD dampers, causing pressure cycles. These pressure cycles are further amplified by the location of the pressure transmitter for each FD damper. The pressure transmitters are located on a common arch; thus a change in left pressure will cause the right pressure to change, and vice versa. The ID damper regulates the boiler steam pressure via a PI controller.
A boiler is a multiple input, multiple output (MIMO) plant. If a change in one input affects more than one output, the plant is considered as interacting. If a MIMO plant is non-interacting (if one input affects only one output), decentralized manipulated variable-control variables pairing can be used.
There are various techniques to design decentralized controllers for MIMO systems. Robust design of decentralized controllers accounts for off-diagonal dynamics of a plant as uncertainty. Robust decentralized control implemented on an industrial boiler showed improved performance over previous PID control. The relative gain array (RGA) is a measure of interaction for decentralized control. The RGA is also a measure of the plant model sensitivity to input uncertainty, such as actuator dynamics. To achieve an efficient operating plant and realize its monetary benefits, it is crucial to identify key performance indicators (KPIs) and set target numbers. For boiler performance monitoring, KPIs were defined so as to identify process targets.
The intricate behavior and relationships associated with boilers present widespread challenges in achieving optimal performance. Regulatory PI controllers provide acceptable steady state performance. However, APC provides more stability and responsiveness in operation because it can handle interaction.
An APC system was installed to optimize the combustion system at the chemical company. The system included a GenSym G2 expert system toolkit developed for Anglo Platinum (referred to as the Anglo Platinum Expert Toolkit (APET)), and AspenTech’s DMCplus model predictive controller.
The real-time expert system, APET, is an object-oriented environment that includes a representation of plant equipment in a knowledge base. All plant variable values are continually updated via OPC communications to the plant control system. APET forms the platform for the APC solution, interfacing between the plant control system and DMCplus controller. The APET system used in this case was configured to continuously monitor the communication links and automatically revive the communication when a failure occurs. APET can include KPI equations. These features were made into an offline reporting tool. APET has many advantages, including real-time calculations for the boiler APC.
An APC system was installed to optimize the chemical company’s combustion system. In this case, the APC controller consisted of a main steam header controller and individual boiler MPC controllers. The steam header MPC controls the steam header pressure by manipulating the ID damper of each boiler. Each boiler ID damper determines the steam demand of the boiler. The change in steam demand causes the header pressure to change accordingly, and thus, the MPC manipulates the ID dampers to maintain the header pressure. There are seven boilers and one steam header pressure; hence, the MPC is overspecified. Each boiler affects the steam header pressure, which means the main steam header is highly interactive.
To increase steam production, the ID damper was opened. This let through more air, increasing combustion and the production of steam. This causes the coal on the grate to be consumed quicker, in turn causing the burn zone to shift toward the rear of the boiler. The boiler controllers react to changes in the ID damper position by increasing or decreasing the stoker speeds, to maintain a ratio between the ID damper position and the length of the coal bed on the grate. This keeps the fire in the correct position in the boiler.
The coal bunker is not covered; moisture buildup requires the coal to dry and remain on the grate for an extended period of time before combustion. Hence, a bias was added to compensate for such scenarios. These controller calculations were implemented and the output was added as control variables (CVs) to the controller as left and right stoker targets. The boiler MPC manipulated the FD damper pressure setpoints by maximizing FD pressure while ensuring that no flames exit the boiler. The controller had a limited ability to minimize the off-gas percentage of oxygen due to the sticktion present on the FD dampers.
The performance of the boiler house was evaluated over a 10-month period consisting of five months without APC and five months with APC. APC provided significant improvement in the steam header pressure stability, reducing the pressure standard deviation (µ) by 50% (Figure 4). The boiler house steam to coal ratio improved from 9.07 to 9.98 (Figure 4). The average boiler saturation was reduced during APC, although there is no significant difference in the amount of steam produced over the comparison period (Figure 5). A one-way ANOVA test was used to evaluate significant differences in PID control versus APC control.
Less coal used
The implementation of an MPC with the APET and DMCplus expert systems stabilized the steam header pressure. This resulted in overall improvement in the boiler operation; most importantly, coal was saved. Automatic communication revival allows for ease of maintenance and increased APC run time. The continuous reports generated readily provide personnel with information to further optimize assets. The replacement of the current FD dampers to a system that is less prone to sticktion will allow the MPC controller to provide more benefits by minimizing oxygen in the off-gas.
Prevlen Rambalee is control engineer in the control and instrumentation department, Anglo Platinum; Gustaf Gous is principal consultant at BluESP; P.G.R. de Villiers is lead process control engineer at the control and instrumentation department, Anglo Platinum; N. McCulloch is process metallurgist at Base Metals Refinery, Anglo Platinum; and G. Humphries, head of technology in the control and instrumentation department, Anglo Platinum.
For more help with process control, see the Control Engineering Process Control Channel at
Also from Control Engineering, read:
Advanced Process Control: Fuzzy Logic and Expert Systems
Driving Plant Optimization with Advanced Process Control
Model-based advanced process control helps create energy efficient plants
– Edited by Mark T. Hoske, Control Engineering, www.controleng.com; posted by Kelsey Kirkley, CFE Media.