Economic Dispatch Using Dynamic Programming
Economic Dispatch Using Dynamic Programming
Dynamic Programming
• If non‐convex input‐output curves are used, we
cannot use an equal incremental cost methodology.
• There are multiple values of MW output for any
given value of incremental cost for non convex
curves.
• For such cases, optimum dispatch is found using
dynamic programming (DP).
• The dynamic programming solution to economic
dispatch is done as an allocation problem.
Dynamic‐Programming (DP)
• DP techniques greatly reduce the computational
effort in finding optimal trajectories or control
policies.
• The theoretical mathematical background, based on
the calculus of variations, is somewhat difficult.
• In the scheduling of power generation systems, DP
techniques have been developed for the following.
• The economic dispatch of thermal systems.
• The solution of hydrothermal economic‐scheduling
problems.
• The practical solution of the unit commitment problem.
1
ED Using Dynamic Programming
Dynamic‐Programming (DP) – Example
• cost of transporting a
unit shipment from
node A to node N.
Dynamic‐Programming (DP) – Example
•Maximum cost path from A to N (ABEHLN)
•The maximum cost to N, starting from any
node on the original maximal path, is
contained in that original path.
•The optimum sequence is called the optimal
policy; any subsequence is a sub policy.
•An optimal policy must contain only optimal
sub policies.
2
ED Using Dynamic Programming
Dynamic‐Programming (DP) – Example
•Bellman and Dreyfus call it the “Principle of
optimality”
•A policy is optimal if, at a stated
stage, whatever the preceding
decisions may have been, the
decisions still to be taken constitute
an optimal policy when the result of
the previous decisions is included.
3
ED Using Dynamic Programming
Dynamic‐Programming (DP) – Example
• Find minimum cost path ?
• There are stages , , , , .
• At the terminus of each stage, there is a set of choices of
nodes to be chosen.
• For example , , ,
• The symbol , represents the “cost” of traversing
stage a( = I,. . . , V) and depends on the variables selected
from the sets { } and { }.
• is the minimum cost for stages I through a to arrive at
some particular node at the end of that stage, starting
from A.
Dynamic‐Programming (DP) – Example
The picture can't be display ed.
4
ED Using Dynamic Programming
Dynamic‐Programming (DP) – Example
Dynamic‐Programming (DP) – Example
5
ED Using Dynamic Programming
Dynamic‐Programming (DP) – Example
Dynamic‐Programming (DP) – Example
6
ED Using Dynamic Programming
Dynamic‐Programming (DP) – Example 2
Resource Allocation Problem
Resource Allocation Problem
• Given a limited amount of money to allocate, the problem is to find
the optimal investment allocation. The only restriction is that
investments must be made in integer amounts.
7
ED Using Dynamic Programming
Resource Allocation Problem
8
ED Using Dynamic Programming
9
ED Using Dynamic Programming
Example – 3E
• Three Units in the System with a total demand D = 310 MW.
• Input – Output characteristics are not smooth nor covex.
10
ED Using Dynamic Programming
11
ED Using Dynamic Programming
Interpolate results for D = 310.
Marginal Unit is Unit 2 so fix it at 110 MW. The results at 110 MW for Unit 2 are
Economic Dispatch and Dynamic Programming
•Poor control performance of the generators
•Automatic Generation Control (AGC)
• Small load addition will result in increasing the
output of appropriate units so that new
generation meets the demand.
• Units must move to new generation value within
a short period of time.
• Large generators don’t change generation output
that easily. There is a maximum rate limit.
12
ED Using Dynamic Programming
Ramp Limit Constraints and Dynamic
Programming
• Small changes in load may result in changes of
output in a number of generators for economic
dispatch.
• Control Action to change the generation levels may
not be possible and probably violate the ramp rates
of generators
• Possible solution is to add ramp rate limits to
economic dispatch formulation
• Short range load forecast is also needed.
Problem Formulation with Ramp Rates
• Given a load to be supplied at time increments t = 1 . . . tmax with load
levels of Ptload and N generators on‐line to supply the load:
13
ED Using Dynamic Programming
Problem Formulation with ramp rates
We must schedule the units to minimize the cost to deliver power over
the time period as:
Base Point and Participation Factors
• Economic Dispatch Problem is solved repeatedly for
different loads.
• Generation levels change from one optimum
schedule to another by a small amount with each
load change.
• A given optimum schedule is called a base point.
• As load changes, a scheduler investigates how much
each generating unit needs to be moved (i.e.
participate) in order for the load to be served at
most economic level.
14
ED Using Dynamic Programming
Base Point and Participation Factors
• For a small change in power output on single unit,
Base Point and Participation Factors
• Let PD be the total demand on the generators (where
PD = Pload + Ploss), then
• the participation factor for each unit are as follows.
15
ED Using Dynamic Programming
Example 3I
16
ED Using Dynamic Programming
Example 3I
17
ED Using Dynamic Programming
Example 3I
Dispatch of thermal generator units with
transmission losses considered
Ngen
Pload Ploss Pi 0
i1
20
18
ED Using Dynamic Programming
Lagrange function with
losses
or
37
Incremental losses
• During the economic dispatch we first determine the
incremental Losses, then consider them fixed, then
perform the economic dispatch Then recalculate the
incremental losses.
• Incremental losses may be calculated by approximate
methods such as the B matrix, or calculate exactly
using a power flow algorithm (OPF).
38
19
ED Using Dynamic Programming
The concept of locational marginal price
(LMP)
39
Generator linear segment cost functions
40
20
ED Using Dynamic Programming
Base dispatch with no line flow limit
LMP at both buses
Generation dispatch: Is 7.5
41
Dispatch with 100 MW line flow limit
Generation dispatch:
LMP at bus 1 is 7.5
Segment MW Price LMP at bus 2 is 8.0
A 400 5.00
C 200 6.50
B 200 7.50 Gen 2 segment D
D 100 8.00
42
21
ED Using Dynamic Programming
Summary
• Economic dispatch is performed at even time
intervals or upon a significant load change in
real time during power system operations
• Any time a study is to be done, some form of
economic dispatch must be performed to
match generation output to load and to get
the approximate minimum cost
43
22