Application of Multivariable Model Prediction Control to Ultra-supercritical Unit - Intelligent Computing for Sustainable Energy and Environment - page 501

2
Q +

2
R +

2
V

min
ΔU k

J k = W k − Y k ΔU k U k − U IRV,k

(6)

s.t.CΔU k ≤ b,

After the extreme value of optimized index J is calculated, it can be seen that
the control variable at the current k moment is:

ΔU ( k )=( A T QA + R ) − 1 A T Q [ W ( k ) −Y 0 ( k )] (7)

in which A is the dynamic matrix of unit predictive model, W the setpoint value
vector, Q the deviation weight matrix and R the control weight matrix. Deviation
weight matrix Q represents the error control degree in P future time domains
and control matrix R characterizes the constraint degree for control increments.
In actual, control strategies are usually made from the whole economic indicator
of the system (comprehensively considering the output and control energy, the
production materials and savings). The combination of steady-state optimiza-
tion and dynamic optimization of the system is realized in [19,20] by overall
linear optimization and rolling optimization conducted by changeable constraint
control. Usually on the basis of equation (3), the manipulated variables of k in-
stant exerted by predictive control on the system is the first group manipulated
variables of the optimization. In the ultra-supercritical fossil power plant pre-
dictive control, the coal feed flow, feed water flow and valve opening bias value
calculated through predictive control algorithm are added to the coal master
control of the coordinated control loop, and along with the correction of feed
water master control and turbine master control on dynamic feedforward loop,
the ideal control effects are therefore realized.

5 The Implementation of Model-Prediction Predictive
Control Scheme

The operation of the control scheme is divided into two phases which are the
simulation phase and the field conduction phase.

a) Test in simulation system: with noise added in the system, the load varied
(referring to the track) and set value changed in order to test reliability of the
algorithm and the software.
b) Test in field: Comparing simulation, controlled variables are sampled from
plant and the algorithm is implemented based on actual set value of the setpoints.
The error should be calculated and curve plotted for synthetical assessment.
Under this stage, the setting of the operating points does not enter the actual
system.
c) Operation of actual system: After finishing the two tests and assessing
results strictly, the control system is put into real operation by setting real
setpoint and manipulated variable into Real-time environment.