Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

restore ff code #13

Merged
merged 5 commits into from
Oct 25, 2023
Merged
Changes from 1 commit
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
152 changes: 103 additions & 49 deletions src/core/feedforward.jl
Original file line number Diff line number Diff line change
Expand Up @@ -83,58 +83,112 @@ function PSI.add_feedforward_constraints!(
return
end

#=
# TODO: Check if this is should be restored
function PSI._add_variable_cost_to_objective!(
container::PSI.OptimizationContainer,
::T,
component::U,
op_cost::PSY.MarketBidCost,
::V,
) where {T <: PSI.EnergySurplusVariable, U <: PSY.Storage, V <: EnergyValueCurve}
component_name = PSY.get_name(component)
@debug "Market Bid" _group = PSI.LOG_GROUP_COST_FUNCTIONS component_name
time_steps = PSI.get_time_steps(container)
initial_time = PSI.get_initial_time(container)
variable_cost_forecast = PSY.get_variable_cost(
component,
op_cost;
start_time=initial_time,
len=length(time_steps),
)
variable_cost_forecast_values = TimeSeries.values(variable_cost_forecast)
parameter_container = PSI._get_cost_function_parameter_container(
container,
PSI.CostFunctionParameter(),
component,
T(),
V(),
eltype(variable_cost_forecast_values),
)
pwl_cost_expressions =
PSI._add_pwl_term!(container, component, variable_cost_forecast_values, T(), V())
jump_model = PSI.get_jump_model(container)
for t in time_steps
PSI.set_parameter!(
parameter_container,
jump_model,
PSY.get_cost(variable_cost_forecast_values[t]),
# Using 1.0 here since we want to reuse the existing code that adds the mulitpler
# of base power times the time delta.
1.0,
component_name,
t,
"""
Adds a constraint to limit the sum of a variable over the number of periods to the source value
"""
struct EnergyLimitFeedforward <: PSI.AbstractAffectFeedforward
optimization_container_key::PSI.OptimizationContainerKey
affected_values::Vector{<:PSI.OptimizationContainerKey}
number_of_periods::Int
function EnergyLimitFeedforward(;
component_type::Type{<:PSY.Component},
source::Type{T},
affected_values::Vector{DataType},
number_of_periods::Int,
meta = CONTAINER_KEY_EMPTY_META,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[JuliaFormatter] reported by reviewdog 🐶

Suggested change
meta = CONTAINER_KEY_EMPTY_META,
meta=CONTAINER_KEY_EMPTY_META,

) where {T}
values_vector = Vector{VariableKey}(undef, length(affected_values))
for (ix, v) in enumerate(affected_values)
if v <: VariableType
values_vector[ix] =
PSI.get_optimization_container_key(v(), component_type, meta)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[JuliaFormatter] reported by reviewdog 🐶

Suggested change
PSI.get_optimization_container_key(v(), component_type, meta)
PSI.get_optimization_container_key(v(), component_type, meta)

else
error(
"EnergyLimitFeedforward is only compatible with VariableType or ParamterType affected values",
)
end
end
new(
PSI.get_optimization_container_key(T(), component_type, meta),
values_vector,
number_of_periods,
)
PSI.add_to_expression!(
end
end

PSI.get_default_parameter_type(::EnergyLimitFeedforward, _) = EnergyLimitParameter
PSI.get_optimization_container_key(ff) = ff.optimization_container_key
PSI.get_number_of_periods(ff) = ff.number_of_periods

@doc raw"""
add_feedforward_constraints(container::OptimizationContainer,
cons_name::Symbol,
param_reference,
var_key::VariableKey)

Constructs a parameterized integral limit constraint to implement feedforward from other models.
The Parameters are initialized using the upper boundary values of the provided variables.


``` sum(variable[var_name, t] for t in 1:affected_periods)/affected_periods <= param_reference[var_name] ```

# LaTeX

`` \sum_{t} x \leq param^{max}``

# Arguments
* container::OptimizationContainer : the optimization_container model built in PowerSimulations
* model::DeviceModel : the device model
* devices::IS.FlattenIteratorWrapper{T} : list of devices
* ff::FixValueFeedforward : a instance of the FixValue Feedforward
"""
function add_feedforward_constraints!(
container::OptimizationContainer,
::DeviceModel,
devices::IS.FlattenIteratorWrapper{T},
ff::EnergyLimitFeedforward,
) where {T <: PSY.Component}
time_steps = get_time_steps(container)
parameter_type = get_default_parameter_type(ff, T)
param = get_parameter_array(container, parameter_type(), T)
multiplier = get_parameter_multiplier_array(container, parameter_type(), T)
affected_periods = get_number_of_periods(ff)
for var in get_affected_values(ff)
variable = get_variable(container, var)
set_name, set_time = JuMP.axes(variable)
IS.@assert_op set_name == [PSY.get_name(d) for d in devices]
IS.@assert_op set_time == time_steps

if affected_periods > set_time[end]
error(
"The number of affected periods $affected_periods is larger than the periods available $(set_time[end])",
)
end
no_trenches = set_time[end] ÷ affected_periods
var_type = get_entry_type(var)
con_ub = add_constraints_container!(
container,
PSI.ProductionCostExpression,
pwl_cost_expressions[t],
component,
t,
FeedforwardIntegralLimitConstraint(),
T,
set_name,
1:no_trenches;
meta = "$(var_type)integral",
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[JuliaFormatter] reported by reviewdog 🐶

Suggested change
meta = "$(var_type)integral",
meta="$(var_type)integral",

)
PSI.add_to_objective_variant_expression!(container, pwl_cost_expressions[t])
end

for name in set_name, i in 1:no_trenches
con_ub[name, i] = JuMP.@constraint(
container.JuMPmodel,
sum(
variable[name, t] for
t in (1 + (i - 1) * affected_periods):(i * affected_periods)
) <= sum(
param[name, t] * multiplier[name, t] for
t in (1 + (i - 1) * affected_periods):(i * affected_periods)
)
)
end
end
return
end
=#

# TODO: It also needs the add parameters code
Loading