- Controller or/and control strategy that is different from conventional PID and/or conventional control strategy - Uses model based controller and modelling system instead of PID controller and closed loop control system - Example : Adaptive PID based o Discrete time o Feedforward controller o Feedforward-feedbackward controller - Model Based o Model Predictive controller (MPC) o Generic Model Controller GMC - Artificial intelligence based o Neural network o Fuzzy logic 2. What are the advantages of using advance control 1. To process variability and non optimal control performance 2. To reduce the variability and increase quality and throughput 3. To locate hidden variability and under performing control loops 4. To achieve more precise control 5. To reduce energy consumption 6. To improve sustainability of a process 3. What is the condition necessary to implement GMC method on a particular system. (Give an example) 1. The application of the GMC is limited to systems which are linear in control (relative degree one). LIMITATION 2. Reasonable process models and parameters. 3. Measurement of controlled variables. 4. Hierarchy of MPC
What is constraints? Give examples
Physical limitation of component in a system(Pressure, Temprature and Quality Specification)
What 1. 2. 3. 4.
are the advantages of MPC
Explicit handling input and state variables Multivariables formation Relatively simple implementation Allow operation closer to the constraint compared to the conventional method and frequently lead to a more profitable operation 5. Relatively low control update rate.
How MPC works?
At time t, solve the optimisation control problem with X(t)=Xo Apply the optimal control action, Uo=U(t) At time = t+1, i) obtain new meaurement ii) Solve X(t+1)=Xo Keep Repeating Properties of MPC 1. Employ a model to formulate a control problem in the space-state representation 2. Able to predict the future and the length of prediction determined by the prediction horizon. 3. Soltion of MPC is subject to contraints and it handles them explicitly 4. Solution obtained is by minimizing the cost Junction J over the prediction horizon. MPC Tuning, Tuning parameters available in MPC are, 1) Prediction Horizon, N Long prediction is desired, however it will accompany by a higher computational complexity due to increased number of decision variables. 2) Weighing matrices Q&R These matrixes affect the closed-loop performance directy. 3) Tuning of state observer Full state measurement is not available. 4) Using software i.e. MATHLAB MPC Graphical user interface.
GMC (Generic Model Control)
- Model Based control system
Example:
Use linear/non-linear model to compute control action
Desired response can be obtained from two tuning parameters (k1 and K2) ADVANTAGE Directly use non linear model of a process to determine a control action. Non-linear models do not need to be linearized. LIMITATION Limited to systems which are linear in control (Relatively degree one) REQUIREMENT Reasonable process models and parameters Measurements of controlled variables