Originally Contributed by: Clayton Barrows
PowerSimulations.jl supports non-linear AC optimal power flow through a deep integration with PowerModels.jl. This example shows a single multi-period optimization of economic dispatch with a full representation of AC optimal power flow.
using SIIPExamples
using PowerSystems
using PowerSimulations
using PowerSystemCaseBuilder
We can use the a TAMU synthetic ERCOT dataset.
The TAMU data format relies on a folder containing .m or .raw files and .csv
files for the time series data. We have provided a parser for the TAMU data format with
the TamuSystem() function. A version of the system with only one week of time series
is included in PowerSystemCaseBuilder.jl, we can use that version here:
sys = build_system(PSYTestSystems, "tamu_ACTIVSg2000_sys")
transform_single_time_series!(sys, 2, Hour(1))
Since we'll be doing non-linear optimization, we need a solver that supports non-linear problems. Ipopt is quite good.
using Ipopt
solver = optimizer_with_attributes(Ipopt.Optimizer)
In the OperationsProblem example we defined a unit-commitment problem with a copper plate representation of the network. Here, we want do define an economic dispatch (linear generation decisions) with an ACOPF network representation. So, starting with the network, we can select from almost any of the endpoints on this tree:
print_tree(PowerSimulations.PM.AbstractPowerModel)
For now, let's just choose a standard ACOPF formulation.
ed_template = OperationsProblemTemplate(QCLSPowerModel)
set_device_model!(ed_template, ThermalStandard, ThermalDispatch)
set_device_model!(ed_template, PowerLoad, StaticPowerLoad)
#set_device_model!(ed_template, FixedAdmittance, StaticPowerLoad) #TODO add constructor for shunts in PSI
Now we can build a 4-hour economic dispatch / ACOPF problem with the TAMU data.
problem = OperationsProblem(
ed_template,
sys,
horizon = 1,
optimizer = solver,
balance_slack_variables = true,
)
build!(problem, output_dir = mktempdir())
And solve it ... (it's infeasible)
solve!(problem)
This notebook was generated using Literate.jl.