Keywords (Extracted from title, table of contents and
abstract of thesis) Transmission, Combinatorial,
Control, Commitment, Optimization, Approaches, Amount, Peak, Unit,
Hybrid, Functions, Forecasted |
Abstract Unit Commitment is
an important and vital optimization task in a power control
centre.After load forecasting it is the second step in the planning
process. It consists of two linked optimization problems. It
comprises unit on/off scheduling problem and the economic dispatch
sub-problem.The on/off scheduling problem is a 0-1 combinatorial
problem with equality and inequality constraints, while the economic
dispatch sub-problem is a nonlinear constrained optimization
problem.
Unit commitment is a nonlinear, large scale, combinatorial,
constrained optimization problem.The complete unit commitment
optimization problem is to minimize the total production cost (TPC)
of utility in such a way that the constraints such as load demand,
spinning reserve, minimum and maximum power limits of units, minimum
up (MUT) and minimum down times (MDT) are satisfied.Therefore, based
on the forecasted load demand, preparing proper on/off schedule of
generators can result in cost saving for utility.It is much more
difficult problem to solve due to its high dimensionality.
The present work is based on the scheduling of thermal units. The
generation of initial feasible UC schedules is much important, for
the UCP. When initial feasible schedules (generation > load +
spinning reserve) are generated randomly, it is difficult to get
feasible schedules for the whole daily forecasted load curve.
In this work initial schedule is generated by considering the peak,
off peak load of the forecasted load curve, must run and must out
units based on a new priority list method.The proposed method is
very efficient and fast in generating initial unit commitment
schedules. In this work the MUT and MDT constraints are checked and
repaired by using bit change operator.The trial solutions were
generated by taking upper four units in the priority list at each
hour to avoid entrapment in local minimum.
In the unit commitment problem, the economic load dispatch (ELD)
sub-problem is an intensive part and its calculations consume a
large amount of time. Convex economic dispatch using load to
efficient unit and incremental cost criterion methods have been
solved. In this work, the ELD calculation for non convex problem has
been solved using genetic algorithm (GA) based on real power search
method.
In the present work, three hybrid approaches have been developed for
the convex and non-convex cost functions and applied first time to
solve the unit commitment problem. To implement these algorithms a
flexible and extensible computational framework has been developed
to run in visual C++ environment. The proposed algorithms are (i)
“hybrid of dynamic programming, particle swarm optimization and
artificial neural network algorithms (DP-PSOANN) for convex cost
function, (ii) “Neuro-Genetic hybrid approach for non-convex cost
function” and (iii) “hybrid of full load average production cost and
maximum power output, for convex and non convex cost functions”. For
comparison the neural network trained with back propagation learning
rule has also been developed. The proposed models have been tested
onI EEE 3 and 10 units standard test systems.The significant
improvement in the total productioncost shows the promise of these
hybrid models.
National utility system, National Transmission and despatch Company
(NTDC) has been reviewed with reference to its operation problems.
Four test systems consisting of 12, 15, 25 and 34 units of NTDC
system have been tested. |