FeedForward Formulations

FeedForwards are the mechanism to define how information is shared between models. Specifically, a FeedForward defines what to do with information passed with an inter-stage chronology in a Simulation. The most common FeedForward is the SemiContinuousFeedForward that affects the semi-continuous range constraints of thermal generators in the economic dispatch problems based on the value of the (already solved) unit-commitment variables.

The creation of a FeedForward requires at least specifying the component_type on which the FeedForward will be applied. The source variable specifies which variable will be taken from the problem solved, for example, the commitment variable of the thermal unit in the unit commitment problem. Finally, the affected_values specify which variables will be affected in the problem to be solved, for example, the next economic dispatch problem.

Table of contents

1. SemiContinuousFeedforward
2. FixValueFeedforward
3. UpperBoundFeedforward
4. LowerBoundFeedforward

SemiContinuousFeedforward

SemiContinuousFeedforward(
    component_type::Type{<:PSY.Component},
    source::Type{T},
    affected_values::Vector{DataType},
    meta = CONTAINER_KEY_EMPTY_META
) where {T}

Variables:

No variables are created

Parameters:

• $\text{on}^\text{th}$ = OnStatusParameter obtained from the source variable, typically the commitment variable of the unit commitment problem $u^\text{th}$.

Objective:

No changes to the objective function.

Expressions:

Adds $-\text{on}^\text{th}P^\text{th,max}$ to the ActivePowerRangeExpressionUB expression and $-\text{on}^\text{th}P^\text{th,min}$ to the ActivePowerRangeExpressionLB expression.

Constraints:

Limits the ActivePowerRangeExpressionUB and ActivePowerRangeExpressionLB by zero as:

\[\begin{align*} & \text{ActivePowerRangeExpressionUB}_t := p_t^\text{th} - \text{on}_t^\text{th}P^\text{th,max} \le 0, \quad \forall t\in \{1, \dots, T\} \\ & \text{ActivePowerRangeExpressionLB}_t := p_t^\text{th} - \text{on}_t^\text{th}P^\text{th,min} \ge 0, \quad \forall t\in \{1, \dots, T\} \end{align*}\]

Thus, if the commitment parameter is zero, the dispatch is limited to zero, forcing to turn off the generator without introducing binary variables in the economic dispatch problem.

FixValueFeedforward

FixValueFeedforward(
    component_type::Type{<:PSY.Component},
    source::Type{T},
    affected_values::Vector{DataType},
    meta = CONTAINER_KEY_EMPTY_META
) where {T}

Variables:

No variables are created

Parameters:

The parameter FixValueParameter is used to match the result obtained from the source variable (from the simulation state).

Objective:

No changes to the objective function.

Expressions:

No changes on expressions.

Constraints:

Set the VariableType from the affected_values to be equal to the source parameter store in FixValueParameter

\[\begin{align*} & \text{AffectedVariable}_t = \text{SourceVariableParameter}_t, \quad \forall t \in \{1,\dots, T\} \end{align*}\]

UpperBoundFeedforward

UpperBoundFeedforward(
    component_type::Type{<:PSY.Component},
    source::Type{T},
    affected_values::Vector{DataType},
    add_slacks::Bool = false,
    meta = CONTAINER_KEY_EMPTY_META
) where {T}

Variables:

If slack variables are enabled:

• UpperBoundFeedForwardSlack
  • Bounds: [0.0, ]
  • Default proportional cost: 1e6
  • Symbol: $p^\text{ff,ubsl}$

Parameters:

The parameter UpperBoundValueParameter stores the result obtained from the source variable (from the simulation state) that will be used as an upper bound to the affected variable.

Objective:

The slack variable is added to the objective function using its large default cost $+ p^\text{ff,ubsl} \cdot 10^6$

Expressions:

No changes on expressions.

Constraints:

Set the VariableType from the affected_values to be lower than the source parameter store in UpperBoundValueParameter.

\[\begin{align*} & \text{AffectedVariable}_t - p_t^\text{ff,ubsl} \le \text{SourceVariableParameter}_t, \quad \forall t \in \{1,\dots, T\} \end{align*}\]

LowerBoundFeedforward

LowerBoundFeedforward(
    component_type::Type{<:PSY.Component},
    source::Type{T},
    affected_values::Vector{DataType},
    add_slacks::Bool = false,
    meta = CONTAINER_KEY_EMPTY_META
) where {T}

Variables:

If slack variables are enabled:

• LowerBoundFeedForwardSlack
  • Bounds: [0.0, ]
  • Default proportional cost: 1e6
  • Symbol: $p^\text{ff,lbsl}$

Parameters:

The parameter LowerBoundValueParameter stores the result obtained from the source variable (from the simulation state) that will be used as a lower bound to the affected variable.

Objective:

The slack variable is added to the objective function using its large default cost $+ p^\text{ff,lbsl} \cdot 10^6$

Expressions:

No changes on expressions.

Constraints:

Set the VariableType from the affected_values to be greater than the source parameter store in LowerBoundValueParameter.

\[\begin{align*} & \text{AffectedVariable}_t + p_t^\text{ff,lbsl} \ge \text{SourceVariableParameter}_t, \quad \forall t \in \{1,\dots, T\} \end{align*}\]