Information Technology Reference
In-Depth Information
⎡
⎤
ΔB
M
(
k
)
ΔQ
1
M
(
k
)
ΔQ
2
M
(
k
)
⎣
⎦
,
ΔB
M
(
k
) is the change to the amount of
Where
Δu
M
(
k
)=
coal(t/h),
ΔQ
1
M
(
k
) is the change to the primary air (Nm
3
/h)
ΔQ
2
M
(
k
)isthe
change of the secondary air (Nm
3
/h). Denote
y
N
1
(
k
)asthe
k
+1 system output
predicted at time
k
.
e
(
k
+1) is error between the actual output and the predicted
output at time
k
+ 1. Then the modified vector of the predictive output is
y
cor
(
k
+1)=
y
N
1
(
k
)+
He
(
k
+1).
2.3 Industrial RTO
In large-scale industrial process control, RTO is usually used together with pre-
dictive control. RTO optimizes operating points based on economic objective of
the process, while MPC implements control based on the operating point opti-
mized by RTO. According to [13], there are several kinds of expression of RTO,
among which one widely used RTO used in practical applications. the RTO with
linear objective function is expressed as the following optimization problem:
min
c
y
y
TG
+
c
u
u
TG
s.t.
u
T
min
≤
u
TG
u
T
max
,
≤
(2)
y
T
min
≤
y
TG
y
T
max
,
≤
y
TG
=
f
(
y,u
)
where
y
TG
is real-time optimization relative vector from output, and
u
TG
is
real-time optimization relative vector from input. The optimal operation point
is achieved via linear optimization of RTO layer.
For the implementation of RTO, there are two general ways: one is construct-
ing a dedicated RTO layer, optimizing the steady-state value for fast systems;
another is combining the RTO layer and MPC to achieve the dual purposes of
optimal control and operation point (see detail in [14]). As shown in [15], direct
combination of RTO and MPC simplifies the control structure.
3 The Optimal Control of Coal-Saving of CFB Boiler
(RTO/DMC)
For the optimal control of the CFB boiler, MPC control can achieve good con-
trol performance, but contribute very little on coal-saving. On the other hand,
operation point optimization by RTO needs the additional devices and com-
munication, which increase the cost of system. Hence, we proposed a control
structure shown in Fig. 2, which embeds the RTO into the DMC controller as
in [15]. The control structure integration coal-saving and dynamic optimization
and make it possible to set the CFB boiler online according to the practical
conditions. Based on the integration of coal and dynamic control optimization
shown in Fig. 2, the predictive control optimization problem is formulated as
Search WWH ::
Custom Search