Model Predictive Control in Power Generation

Emerson's Ranjit RaoModel predictive control in power generation was the subject of a 2017 Ovation Users Group presentation by Emerson’s Ranjit Rao. Rajit opened comparing classical control with model-based control. In classical control, a proportional-derivative-integral loop is used with feedback. There is a setpoint, the actual process variable and the error or difference between the process variable and setpoint. Based on the dynamics of the process and load disturbances, the amount of proportional, derivative and integral gain is set to have to have the loop control the process effectively.

The PID approximates the process and does not handle higher order responses well. Model-based control is ideal for processes with long dead-times and higher order responses. A built-in optimizer provides the ability to focus on an outcome such as profitability. The model-based controllers can model dynamic disturbances. In the Ovation system, these model predictive control (MPC) algorithms run in the Ovation controller.

 Ranjit described internal model control as a process that simulates the response of the system in order to estimate the outcome of a system disturbance. It is inherently more robust than classical control. In model based control the deviation between the actual process variable with the model process variable. The difference is compared with where the setpoint wants the process variable to be.

Model predictive control is based on an iterative, finite-horizon optimization of a plant model. The horizon is recalculated with every scan and continuously updates.

The Ovation system has dynamic matrix control (DMC) and model predictive controller (MPC) and are part of the APC Toolkit. The DMC handles 6 process variables, 6 disturbance variables, 5 manipulated variables and has a prediction horizon of 45 samples. It handles output constraints related to actuator or control action rate and range limits. Steam temperature control is a good application for DMC. Steam temperature is a third order process with dead time. Model-based control decouples steam temperature and boiler load.

Other application for MPC he discussed included selective catalytic reduction (SCR) control & ammonia optimization and unit coordinated control.

The post Model Predictive Control in Power Generation appeared first on the Emerson Process Experts blog.