Page 78 - Mathematical Models and Algorithms for Power System Optimization

68 Chapter 3

Rule 6: If an interval between the maintenance period of some units cannot be extended or if the
operation index is still OPindex>1.0 after extension, then forecast loads in some time intervals
should be curtailed to make the operation index less than 1.0.

3.5.3.3 Rules of verifying area constraint

Rule 7: In the given time intervals, if the scheduled maintenance capacity or the number of
maintenance units in a certain area is over the constraint, then the maintenance schedule of
some units with lower maintenance priority should be shifted.

3.5.3.4 Rules of verifying power plant constraint
Rule 8: In the given time intervals, the maintenance duration of the units within two plants
should not be overlapped. Namely, units in one power plant and those in another shall be
staggered in terms of maintenance scheduling.

Rule 9: If the two units within the same plant have the overlapping maintenance duration, then
the maintenance duration of the unit with lower priority should be shifted away.

3.5.3.5 Rules of verifying manpower constraint

Rule 10: If the maintenance manpower of a unit is not enough, then the maintenance duration of
the unit should be shifted.
The constraints and objectives of the GMS problem can be described by fuzzy membership
functions, such as fuzzy maintenance window constraint, fuzzy constraint of area maintenance
capacity, fuzzy simultaneous maintenance constraint, and fuzzy manpower constraint, as well
as fuzzy objectives of reserve margin and production costs, etc.

3.6 Calculation Procedure of GMS Optimization

3.6.1 Search Paths and Recursive Formulas of Fuzzy Dynamic Programming

The optimized decision of GMS depends on the intersection of objective function and fuzzy
μ
μ
μ
μ
μ
μ
membership function of constraint boundary e ,eμ g2 and e , e ,e ,e , e i, tÞ is used to
ð
D
c4
c3
c2
c1
g1
express the highest fuzzy membership function value of all maintenance schedule states from
the previous unit to the current unit in the recursive process of FDP, and the optimal decision
μ D Þ of FDP depends on such e i, tÞ.
μ
ð
e optð
D
The search strategy of the FDP is shown in Fig. 3.9.