re-designed the LMP methods

The LMP and AELMP methods are re-designed to be dependent on the instance object instead of input files, and to have a unified API style for purposes of flexibility and consistency.
pull/27/head
Jun He 3 years ago
parent 8fc84412eb
commit 5c91dc2ac9

@ -7,6 +7,7 @@ module UnitCommitment
include("instance/structs.jl") include("instance/structs.jl")
include("model/formulations/base/structs.jl") include("model/formulations/base/structs.jl")
include("solution/structs.jl") include("solution/structs.jl")
include("lmp/structs.jl")
include("model/formulations/ArrCon2000/structs.jl") include("model/formulations/ArrCon2000/structs.jl")
include("model/formulations/CarArr2006/structs.jl") include("model/formulations/CarArr2006/structs.jl")
@ -17,6 +18,8 @@ include("model/formulations/MorLatRam2013/structs.jl")
include("model/formulations/PanGua2016/structs.jl") include("model/formulations/PanGua2016/structs.jl")
include("solution/methods/XavQiuWanThi2019/structs.jl") include("solution/methods/XavQiuWanThi2019/structs.jl")
include("model/formulations/WanHob2016/structs.jl") include("model/formulations/WanHob2016/structs.jl")
include("lmp/lmp/structs.jl")
include("lmp/aelmp/structs.jl")
include("import/egret.jl") include("import/egret.jl")
include("instance/read.jl") include("instance/read.jl")
@ -56,5 +59,7 @@ include("utils/log.jl")
include("utils/benchmark.jl") include("utils/benchmark.jl")
include("validation/repair.jl") include("validation/repair.jl")
include("validation/validate.jl") include("validation/validate.jl")
include("lmp/lmp/compute.jl")
include("lmp/aelmp/compute.jl")
end end

@ -0,0 +1,231 @@
# UnitCommitment.jl: Optimization Package for Security-Constrained Unit Commitment
# Copyright (C) 2020, UChicago Argonne, LLC. All rights reserved.
# Released under the modified BSD license. See COPYING.md for more details.
using JuMP
"""
function compute_lmp(
model::JuMP.Model,
method::AELMP.Method;
optimizer = nothing,
)
Calculates the approximate extended locational marginal prices of the given unit commitment instance.
The AELPM does the following three things:
1. It removes the minimum generation requirement for each generator
2. It averages the start-up cost over the offer blocks for each generator
3. It relaxes all the binary constraints and integrality
Returns a dictionary of AELMPs. Each key is usually a tuple of "Bus name" and time index.
NOTE: this approximation method is not fully developed. The implementation is based on MISO Phase I only.
1. It only supports Fast Start resources. More specifically, the minimum up/down time has to be zero.
2. The method does NOT support time series of start-up costs.
3. The method can only calculate for the first time slot if allow_offline_participation=false.
Arguments
---------
- `model`:
the UnitCommitment model, must be solved before calling this function if offline participation is not allowed.
- `method`:
the AELMP method, must be specified.
- `optimizer`:
the optimizer for solving the LP problem.
Examples
--------
```julia
using UnitCommitment
using Cbc
using HiGHS
import UnitCommitment:
AELMP
# Read benchmark instance
instance = UnitCommitment.read("instance.json")
# Construct model (using state-of-the-art defaults)
model = UnitCommitment.build_model(
instance = instance,
optimizer = Cbc.Optimizer,
variable_names = true,
)
# Get the AELMP with the default policy:
# 1. Offline generators are allowed to participate in pricing
# 2. Start-up costs are considered.
# DO NOT use Cbc as the optimizer here. Cbc does not support dual values.
my_aelmp_default = UnitCommitment.compute_lmp(
model, # pre-solving is optional if allowing offline participation
AELMP.Method(),
optimizer = HiGHS.Optimizer
)
# Get the AELMPs with an alternative policy
# 1. Offline generators are NOT allowed to participate in pricing
# 2. Start-up costs are considered.
# UC model must be solved first if offline generators are NOT allowed
UnitCommitment.optimize!(model)
# then call the AELMP method
my_aelmp_alt = UnitCommitment.compute_lmp(
model, # pre-solving is required here
AELMP.Method(
allow_offline_participation=false,
consider_startup_costs=true
),
optimizer = HiGHS.Optimizer
)
# Accessing the 'my_aelmp_alt' dictionary
# Example: "b1" is the bus name, 1 is the first time slot
@show my_aelmp_alt["b1", 1]
```
"""
function _preset_aelmp_parameters!(
method::AELMP.Method,
model::JuMP.Model
)
# this function corrects the allow_offline_participation parameter to match the model status
# CHECK: model must be solved if allow_offline_participation=false
if method.allow_offline_participation # do nothing
@info "Offline generators are allowed to participate in pricing."
else
if isnothing(model)
@warn "No UC model is detected. A solved UC model is required if allow_offline_participation == false."
@warn "Setting parameter allow_offline_participation = true"
method.allow_offline_participation = true # and do nothing else
elseif !has_values(model)
@warn "The UC model has no solution. A solved UC model is required if allow_offline_participation == false."
@warn "Setting parameter allow_offline_participation = true"
method.allow_offline_participation = true # and do nothing else
else
# the inputs are correct
@info "Offline generators are NOT allowed to participate in pricing."
@info "Offline generators will be removed for the approximation."
end
end
# CHECK: start up cost consideration
if method.consider_startup_costs
@info "Startup costs are considered."
else
@info "Startup costs are NOT considered."
end
end
function _modify_instance!(
instance::UnitCommitmentInstance,
model::JuMP.Model,
method::AELMP.Method
)
# this function modifies the instance units (generators)
# 1. remove (if NOT allowing) the offline generators
if !method.allow_offline_participation
for unit in instance.units
# remove based on the solved UC model result
# here, only look at the first time slot (TIME-SERIES-NOT-SUPPORTED)
if value(model[:is_on][unit.name, 1]) == 0
# unregister from the bus
filter!(x -> x.name != unit.name, unit.bus.units)
# unregister from the reserve
for r in unit.reserves
filter!(x -> x.name != unit.name, r.units)
end
end
end
# unregister the units
filter!(x -> value(model[:is_on][x.name, 1]) != 0, instance.units)
end
for unit in instance.units
# 2. set min generation requirement to 0 by adding 0 to production curve and cost
# min_power & min_costs are vectors with dimension T
if unit.min_power[1] != 0
first_cost_segment = unit.cost_segments[1]
pushfirst!(unit.cost_segments, CostSegment(
ones(size(first_cost_segment.mw)) * unit.min_power[1],
ones(size(first_cost_segment.cost)) * unit.min_power_cost[1] / unit.min_power[1]
))
unit.min_power = zeros(size(first_cost_segment.mw))
unit.min_power_cost = zeros(size(first_cost_segment.cost))
end
# 3. average the start-up costs (if considering)
# for now, consider first element only (TIME-SERIES-NOT-SUPPORTED)
# if consider_startup_costs = false, then use the current first_startup_cost
first_startup_cost = unit.startup_categories[1].cost
if method.consider_startup_costs
additional_unit_cost = first_startup_cost / unit.max_power[1]
for i in eachindex(unit.cost_segments)
unit.cost_segments[i].cost .+= additional_unit_cost
end
first_startup_cost = 0.0 # zero out the start up cost
end
# 4. other adjustments...
### FIXME in the future
# MISO Phase I: can ONLY solve fast-starts, force all startup time to be 0
unit.startup_categories = StartupCategory[StartupCategory(0, first_startup_cost)]
unit.initial_status = -100
unit.initial_power = 0
unit.min_uptime = 0
unit.min_downtime = 0
### END FIXME
end
instance.units_by_name = Dict(g.name => g for g in instance.units)
end
function compute_lmp(
model::JuMP.Model,
method::AELMP.Method;
optimizer = nothing
)
# Error if a linear optimizer is not specified
if isnothing(optimizer)
@error "Please supply a linear optimizer."
return nothing
end
@info "Calculating the AELMP..."
@info "Building the approximation model..."
# get the instance and make a deep copy
instance = deepcopy(model[:instance])
# preset the method to match the model status (solved, unsolved, not supplied)
_preset_aelmp_parameters!(method, model)
# modify the instance (generator)
_modify_instance!(instance, model, method)
# prepare the result dictionary and solve the model
elmp = OrderedDict()
@info "Solving the approximation model."
approx_model = build_model(instance=instance, variable_names=true)
# relax the binary constraint, and relax integrality
for v in all_variables(approx_model)
if is_binary(v)
unset_binary(v)
end
end
relax_integrality(approx_model)
set_optimizer(approx_model, optimizer)
# solve the model
# set_silent(approx_model)
optimize!(approx_model)
# access the dual values
@info "Getting dual values (AELMPs)."
for (key, val) in approx_model[:eq_net_injection]
elmp[key] = dual(val)
end
return elmp
end

@ -0,0 +1,41 @@
# UnitCommitment.jl: Optimization Package for Security-Constrained Unit Commitment
# Copyright (C) 2020, UChicago Argonne, LLC. All rights reserved.
# Released under the modified BSD license. See COPYING.md for more details.
module AELMP
import ..PricingMethod
"""
mutable struct Method
allow_offline_participation::Bool,
consider_startup_costs::Bool
end
------
- `allow_offline_participation`:
defaults to true.
If true, offline assets are allowed to participate in pricing.
- `consider_startup_costs`:
defaults to true.
If true, the start-up costs are averaged over each unit production; otherwise the production costs stay the same.
"""
mutable struct Method <: PricingMethod
allow_offline_participation::Bool
consider_startup_costs::Bool
function Method(;
allow_offline_participation::Bool = true,
consider_startup_costs::Bool = true
)
return new(
allow_offline_participation,
consider_startup_costs
)
end
end
end

@ -0,0 +1,123 @@
# UnitCommitment.jl: Optimization Package for Security-Constrained Unit Commitment
# Copyright (C) 2020, UChicago Argonne, LLC. All rights reserved.
# Released under the modified BSD license. See COPYING.md for more details.
using JuMP
"""
function compute_lmp(
model::JuMP.Model,
method::LMP.Method;
optimizer = nothing
)
Calculates the locational marginal prices of the given unit commitment instance.
Returns a dictionary of LMPs. Each key is usually a tuple of "Bus name" and time index.
Returns nothing if there is an error in solving the LMPs.
Arguments
---------
- `model`:
the UnitCommitment model, must be solved before calling this function.
- `method`:
the LMP method, must be specified.
- `optimizer`:
the optimizer for solving the LP problem.
Examples
--------
```julia
using UnitCommitment
using Cbc
using HiGHS
import UnitCommitment:
LMP
# Read benchmark instance
instance = UnitCommitment.read("instance.json")
# Construct model (using state-of-the-art defaults)
model = UnitCommitment.build_model(
instance = instance,
optimizer = Cbc.Optimizer,
)
# Get the LMPs before solving the UC model
# Error messages will be displayed and the returned value is nothing.
# lmp = UnitCommitment.compute_lmp(model, LMP.Method(), optimizer = HiGHS.Optimizer) # DO NOT RUN
UnitCommitment.optimize!(model)
# Get the LMPs after solving the UC model (the correct way)
# DO NOT use Cbc as the optimizer here. Cbc does not support dual values.
# Compute regular LMP
my_lmp = UnitCommitment.compute_lmp(
model,
LMP.Method(),
optimizer = HiGHS.Optimizer,
)
# Accessing the 'my_lmp' dictionary
# Example: "b1" is the bus name, 1 is the first time slot
@show my_lmp["b1", 1]
```
"""
function compute_lmp(
model::JuMP.Model,
method::LMP.Method;
optimizer = nothing
)
# Error if a linear optimizer is not specified
if isnothing(optimizer)
@error "Please supply a linear optimizer."
return nothing
end
# Validate model, the UC model must be solved beforehand
if !has_values(model)
@error "The UC model must be solved before calculating the LMPs."
@error "The LMPs are NOT calculated."
return nothing
end
# Prepare the LMP result dictionary
lmp = OrderedDict()
# Calculate LMPs
# Fix all binary variables to their optimal values and relax integrality
@info "Calculating LMPs..."
@info "Fixing all binary variables to their optimal values and relax integrality."
vals = Dict(v => value(v) for v in all_variables(model))
for v in all_variables(model)
if is_binary(v)
unset_binary(v)
fix(v, vals[v])
end
end
# fix!(model, model[:solution])
relax_integrality(model)
set_optimizer(model, optimizer)
# Solve the LP
@info "Solving the LP..."
JuMP.optimize!(model)
# Obtain dual values (LMPs) and store into the LMP dictionary
@info "Getting dual values (LMPs)..."
for (key, val) in model[:eq_net_injection]
lmp[key] = dual(val)
end
# Return the LMP dictionary
@info "Calculation completed."
return lmp
end

@ -0,0 +1,18 @@
# UnitCommitment.jl: Optimization Package for Security-Constrained Unit Commitment
# Copyright (C) 2020, UChicago Argonne, LLC. All rights reserved.
# Released under the modified BSD license. See COPYING.md for more details.
"""
Formulation described in:
Arroyo, J. M., & Conejo, A. J. (2000). Optimal response of a thermal unit
to an electricity spot market. IEEE Transactions on power systems, 15(3),
1098-1104. DOI: https://doi.org/10.1109/59.871739
"""
module LMP
import ..PricingMethod
struct Method <: PricingMethod end
end

@ -0,0 +1,5 @@
# UnitCommitment.jl: Optimization Package for Security-Constrained Unit Commitment
# Copyright (C) 2020, UChicago Argonne, LLC. All rights reserved.
# Released under the modified BSD license. See COPYING.md for more details.
abstract type PricingMethod end
Loading…
Cancel
Save