Skip to content

Commit

Permalink
Add configurable scaling before ACOPF solving (#62)
Browse files Browse the repository at this point in the history 
* Add scaling of variables/constraints before ACOPF solving.
* Add json serialization/deserialization of new parameters for scaling.
* Add TUs for Java developments.
* Add configurability of scaling factors in AMPL code.
* Add TUs with comparisons of the outputs of AMPL code.
* Remove local path in config.yml.
* Add documentation for scaling before ACOPF solving.
* Fix TUs.
* Remove acopf_scaling.run file.
* Update README.
* Update OR model in Java.
* Update authors.
* Refactor variable names in Java.
* Refactor variable names in AMPL/README.
* Refactor OpenReacParametersDeserializer class.

---------

Signed-off-by: parvy <pierre.arvy@artelys.com>
  • Loading branch information
@p-arvy
p-arvy authored Apr 16, 2024
1 parent 2f21380 commit 6acab12
Show file tree
Hide file tree
Showing 25 changed files with 358 additions and 91 deletions.
45 changes: 25 additions & 20 deletions open-reac/README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -81,25 +81,29 @@ The user can configure the run with the dedicated Java interface
Specifically, the user can set various parameters and thresholds used in the preprocessing and modeling of the reactive OPF.
These are specified in the file `param_algo.txt`:

| Parameter | Description | Default value | Domain |
|------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------|-----------------------------------------------|
| `log_level_ampl` | Level of display for AMPL prints | INFO | {DEBUG, INFO, WARNING, ERROR} |
| `log_level_knitro` | Level of display for solver prints (see [AMPL documentation](https://dev.ampl.com/ampl/options.html)) | $1$ | {0, 1, 2} |
| `objective_choice` | Choice of the objective function for the ACOPF (see [7](#7-alternative-current-optimal-power-flow)) | $0$ | {0, 1, 2} |
| `ratio_voltage_target` | Ratio to calculate target V of buses when `objective_choice` is set to $1$ (see [7](#7-alternative-current-optimal-power-flow)) | $0.5$ | $\[0; 1\]$ |
| `coeff_alpha` | Weight to favor more/less minimization of active power produced by generators or deviation between them and target values (see [6.2](#62-alternative-current-optimal-power-flow)) | $1$ | $\[0; 1\]$ |
| `Pnull` | Threshold of active and reactive powers considered as null | $0.01$ (MW) | $\[0; 1\]$ |
| `Znull` | Threshold of impedance considered as null (see [4.2](#)) | $10^{-5}$ (p.u.) | $\[0; 0.1\]$ |
| `epsilon_nominal_voltage` | Threshold to ignore voltage levels with nominal voltage lower than it | $1$ (kV) | $\mathbb{R}^{+}$ |
| `min_plausible_low_voltage_limit` | Consistency bound for low voltage limit of voltage levels (see [4.1](#41-voltage-level-limits-computation)) | $0.5$ (p.u.) | $\mathbb{R}^{+}$ |
| `max_plausible_high_voltage_limit` | Consistency bound for high voltage limit of voltage levels (see [4.1](#41-voltage-level-limits-computation)) | $1.5$ (p.u.) | [`min_plausible_low_voltage_limit`; $\infty$] |
| `ignore_voltage_bounds` | Threshold to replace voltage limits of voltage levels with nominal voltage lower than it, by [min_plausible_low_voltage_limit; max_plausible_high_voltage_limit] | $0$ (p.u.) | $\mathbb{R}^{+}$ |
| `buses_with_reactive_slacks` | Choice of which buses will have reactive slacks attached in ACOPF solving (see [7](#7-alternative-current-optimal-power-flow)) | NO_GENERATION | {CONFIGURED, NO_GENERATION, ALL} |
| `PQmax` | Threshold for maximum active and reactive power considered in correction of generator limits (see [4.4](#44-pq-units-domain)) | $9000$ (MW, MVAr) | $\mathbb{R}$ |
| `defaultPmax` | Threshold for correction of high active power limit produced by generators (see [4.4](#44-pq-units-domain)) | $1000$ (MW) | $\mathbb{R}$ |
| `defaultPmin` | Threshold for correction of low active power limit produced by generators (see [4.4](#44-pq-units-domain)) | $0$ (MW) | $\mathbb{R}$ |
| `defaultQmaxPmaxRatio` | Ratio used to calculate threshold for corrections of high/low reactive power limits (see [4.4](#44-pq-units-domain)) | $0.3$ (MVAr/MW) | $\mathbb{R}$ |
| `minimalQPrange` | Threshold to fix active (resp. reactive) power of generators with active (resp. reactive) power limits that are closer than it (see [4.4](#44-pq-units-domain)) | $1$ (MW, MVAr) | $\mathbb{R}$ |
| Parameter | Description | Default value | Domain |
|---------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------|-----------------------------------------------|
| `log_level_ampl` | Level of display for AMPL prints | INFO | {DEBUG, INFO, WARNING, ERROR} |
| `log_level_knitro` | Level of display for solver prints (see [AMPL documentation](https://dev.ampl.com/ampl/options.html)) | $1$ | {0, 1, 2} |
| `objective_choice` | Choice of the objective function for the ACOPF (see [7](#7-alternative-current-optimal-power-flow)) | $0$ | {0, 1, 2} |
| `ratio_voltage_target` | Ratio to calculate target V of buses when `objective_choice` is set to $1$ (see [7](#7-alternative-current-optimal-power-flow)) | $0.5$ | $\[0; 1\]$ |
| `coeff_alpha` | Weight to favor more/less minimization of active power produced by generators or deviation between them and target values (see [6.2](#62-alternative-current-optimal-power-flow)) | $1$ | $\[0; 1\]$ |
| `Pnull` | Threshold of active and reactive powers considered as null | $0.01$ (MW) | $\[0; 1\]$ |
| `Znull` | Threshold of impedance considered as null (see [4.2](#)) | $10^{-5}$ (p.u.) | $\[0; 0.1\]$ |
| `epsilon_nominal_voltage` | Threshold to ignore voltage levels with nominal voltage lower than it | $1$ (kV) | $\mathbb{R}^{+}$ |
| `min_plausible_low_voltage_limit` | Consistency bound for low voltage limit of voltage levels (see [4.1](#41-voltage-level-limits-computation)) | $0.5$ (p.u.) | $\mathbb{R}^{+}$ |
| `max_plausible_high_voltage_limit` | Consistency bound for high voltage limit of voltage levels (see [4.1](#41-voltage-level-limits-computation)) | $1.5$ (p.u.) | [`min_plausible_low_voltage_limit`; $\infty$] |
| `ignore_voltage_bounds` | Threshold to replace voltage limits of voltage levels with nominal voltage lower than it, by [min_plausible_low_voltage_limit; max_plausible_high_voltage_limit] | $0$ (p.u.) | $\mathbb{R}^{+}$ |
| `buses_with_reactive_slacks` | Choice of which buses will have reactive slacks attached in ACOPF solving (see [7](#7-alternative-current-optimal-power-flow)) | NO_GENERATION | {CONFIGURED, NO_GENERATION, ALL} |
| `PQmax` | Threshold for maximum active and reactive power considered in correction of generator limits (see [4.4](#44-pq-units-domain)) | $9000$ (MW, MVAr) | $\mathbb{R}$ |
| `defaultPmax` | Threshold for correction of high active power limit produced by generators (see [4.4](#44-pq-units-domain)) | $1000$ (MW) | $\mathbb{R}$ |
| `defaultPmin` | Threshold for correction of low active power limit produced by generators (see [4.4](#44-pq-units-domain)) | $0$ (MW) | $\mathbb{R}$ |
| `defaultQmaxPmaxRatio` | Ratio used to calculate threshold for corrections of high/low reactive power limits (see [4.4](#44-pq-units-domain)) | $0.3$ (MVAr/MW) | $\mathbb{R}$ |
| `minimalQPrange` | Threshold to fix active (resp. reactive) power of generators with active (resp. reactive) power limits that are closer than it (see [4.4](#44-pq-units-domain)) | $1$ (MW, MVAr) | $\mathbb{R}$ |
| `default_variable_scaling_factor` | Default scaling factor applied to all the variables (except reactive slacks and transformer ratios) before ACOPF solving | $1$ | $\mathbb{R}^{*,+}$ |
| `default_constraint_scaling_factor` | Default scaling factor applied to all the constraints before ACOPF solving | $1$ | $\mathbb{R}^{+}$ |
| `reactive_slack_variable_scaling_factor` | Scaling factor applied to all reactive slacks variables before ACOPF solving (see [7](#7-alternative-current-optimal-power-flow)) | $0.1$ | $\mathbb{R}^{*,+}$ |
| `transformer_ratio_variable_scaling_factor` | Scaling factor applied to all transformer ratio variables before ACOPF solving (see [7](#7-alternative-current-optimal-power-flow)) | $0.001$ | $\mathbb{R}^{*,+}$ |


In addition to the previous parameters, the user can specify which
Expand Down Expand Up @@ -220,7 +224,7 @@ Consequently, **buses connected to the slack only by HVDC lines are excluded**.

This component is determined by solving the following optimization problem (the variables are bolded):

$$\text{minimize} \left(\sum\limits_{i} \boldsymbol{\theta_i^{cc}}\right)$$
$$\text{maximize} \left(\sum\limits_{i} \boldsymbol{\theta_i^{cc}}\right)$$

where $\boldsymbol{\theta_i^{cc}}$ is the voltage angle of bus $i$, and with the following constraints:

Expand Down Expand Up @@ -348,6 +352,7 @@ high coefficient ($10$) to drive it towards $0$, ensuring reactive power balance

Before solving the ACOPF, the voltage magnitudes $\boldsymbol{V_i}$ are warm-started with $V_i^t$
(specified in `ampl_network_buses.txt`), as well as the voltage phases $\boldsymbol{\theta_i}$ with the results of the DCOPF (see [6](#6-direct-current-optimal-power-flow)).
Please also note that a scaling is applied with user-defined values before solving the ACOPF.

The solving is considered as successful if the non-linear solver employed (see [Non-linear solver](#non-linear-optimization-solver))
finds a feasible approximate solution (**even if the sum of
Expand Down
80 changes: 80 additions & 0 deletions open-reac/src/main/java/com/powsybl/openreac/parameters/input/OpenReacParameters.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -58,6 +58,22 @@ public class OpenReacParameters {

private ReactiveSlackBusesMode reactiveSlackBusesMode = ReactiveSlackBusesMode.NO_GENERATION;

private static final String DEFAULT_VARIABLE_SCALING_FACTOR = "default_variable_scaling_factor";

private double defaultVariableScalingFactor = 1;

private static final String DEFAULT_CONSTRAINT_SCALING_FACTOR = "default_constraint_scaling_factor";

private double defaultConstraintScalingFactor = 1;

private static final String REACTIVE_SLACK_VARIABLE_SCALING_FACTOR = "reactive_slack_variable_scaling_factor";

private double reactiveSlackVariableScalingFactor = 1e-1;

private static final String TWO_WINDING_TRANSFORMER_RATIO_VARIABLE_SCALING_FACTOR = "transformer_ratio_variable_scaling_factor";

private double twoWindingTransformerRatioVariableScalingFactor = 1e-3;

/**
* Override some voltage level limits in the network. This will NOT modify the network object.
* <p>
Expand Down Expand Up @@ -216,6 +232,66 @@ public OpenReacParameters setReactiveSlackBusesMode(ReactiveSlackBusesMode react
return this;
}

/**
* @return the default scaling value of all the variables in ACOPF solving.
*/
public double getDefaultVariableScalingFactor() {
return defaultVariableScalingFactor;
}

public OpenReacParameters setDefaultVariableScalingFactor(double defaultVariableScalingFactor) {
if (defaultVariableScalingFactor <= 0 || Double.isNaN(defaultVariableScalingFactor)) {
throw new IllegalArgumentException("Default scaling factor for variables must be > 0 and defined to be consistent.");
}
this.defaultVariableScalingFactor = defaultVariableScalingFactor;
return this;
}

/**
* @return the default scaling value of all the constraints in ACOPF solving.
*/
public double getDefaultConstraintScalingFactor() {
return defaultConstraintScalingFactor;
}

public OpenReacParameters setDefaultConstraintScalingFactor(double defaultConstraintScalingFactor) {
if (defaultConstraintScalingFactor < 0 || Double.isNaN(defaultConstraintScalingFactor)) {
throw new IllegalArgumentException("Default scaling factor for constraints must be >= 0 and defined to be consistent.");
}
this.defaultConstraintScalingFactor = defaultConstraintScalingFactor;
return this;
}

/**
* @return the scaling value of reactive slack variables in ACOPF solving.
*/
public double getReactiveSlackVariableScalingFactor() {
return reactiveSlackVariableScalingFactor;
}

public OpenReacParameters setReactiveSlackVariableScalingFactor(double reactiveSlackVariableScalingFactor) {
if (reactiveSlackVariableScalingFactor <= 0 || Double.isNaN(reactiveSlackVariableScalingFactor)) {
throw new IllegalArgumentException("Scaling factor for reactive slack variables must be > 0 and defined to be consistent.");
}
this.reactiveSlackVariableScalingFactor = reactiveSlackVariableScalingFactor;
return this;
}

/**
* @return the scaling value of transformer ratios in ACOPF solving.
*/
public double getTwoWindingTransformerRatioVariableScalingFactor() {
return twoWindingTransformerRatioVariableScalingFactor;
}

public OpenReacParameters setTwoWindingTransformerRatioVariableScalingFactor(double twoWindingTransformerRatioVariableScalingFactor) {
if (twoWindingTransformerRatioVariableScalingFactor <= 0 || Double.isNaN(twoWindingTransformerRatioVariableScalingFactor)) {
throw new IllegalArgumentException("Scaling factor for transformer ratio variables must be > 0 and defined to be consistent.");
}
this.twoWindingTransformerRatioVariableScalingFactor = twoWindingTransformerRatioVariableScalingFactor;
return this;
}

public List<String> getVariableShuntCompensators() {
return variableShuntCompensators;
}
Expand Down Expand Up @@ -247,6 +323,10 @@ public List<OpenReacAlgoParam> getAllAlgorithmParams() {
allAlgoParams.add(new OpenReacAlgoParamImpl(MIN_PLAUSIBLE_LOW_VOLTAGE_LIMIT_KEY, Double.toString(minPlausibleLowVoltageLimit)));
allAlgoParams.add(new OpenReacAlgoParamImpl(MAX_PLAUSIBLE_HIGH_VOLTAGE_LIMIT_KEY, Double.toString(maxPlausibleHighVoltageLimit)));
allAlgoParams.add(reactiveSlackBusesMode.toParam());
allAlgoParams.add(new OpenReacAlgoParamImpl(DEFAULT_VARIABLE_SCALING_FACTOR, Double.toString(defaultVariableScalingFactor)));
allAlgoParams.add(new OpenReacAlgoParamImpl(DEFAULT_CONSTRAINT_SCALING_FACTOR, Double.toString(defaultConstraintScalingFactor)));
allAlgoParams.add(new OpenReacAlgoParamImpl(REACTIVE_SLACK_VARIABLE_SCALING_FACTOR, Double.toString(reactiveSlackVariableScalingFactor)));
allAlgoParams.add(new OpenReacAlgoParamImpl(TWO_WINDING_TRANSFORMER_RATIO_VARIABLE_SCALING_FACTOR, Double.toString(twoWindingTransformerRatioVariableScalingFactor)));
return allAlgoParams;
}

Expand Down
Loading

0 comments on commit 6acab12

Please sign in to comment.