State Estimation
For further information on this topic, please see the PMU State Estimation or DC State Estimation sections of the Manual. Below, we have provided a list of functions that can be utilized for power flow analysis.
Observability Analysis
AC State Estimation
PMU State Estimation
DC State Estimation
Bad Data Analysis
To load state estimation API functionalities into the current scope, utilize the following command:
using JuliaGrid
Observability Analysis
JuliaGrid.islandTopologicalFlow
— MethodislandTopologicalFlow(system::PowerSystem, device::Measurement)
The function utilizes a topological approach to detect flow observable islands, resulting in the formation of disconnected and loop-free subgraphs. It is assumed that active and reactive power measurements are paired, indicating a standard observability analysis. In this analysis, islands formed by active power measurements correspond to those formed by reactive power measurements.
Arguments
To define flow observable islands, this function necessitates the composite types PowerSystem
and Measurement
.
Returns
The function returns an type Island
, containing information about the islands:
island
: a list enumerating observable islands with indices of buses;bus
: the positions of buses in relation to each island;tie
: tie data associated with buses and branches.
Examples
Find flow islands for the provided set of measurements:
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
statusWattmeter!(system, device; inservice = 15)
device.varmeter.reactive.status = copy(device.wattmeter.active.status)
islands = islandTopologicalFlow(system, device)
JuliaGrid.islandTopological
— MethodislandTopological(system::PowerSystem, meter::Wattmeter)
The function employs a topological method to identify maximal observable islands. Specifically, it employs measurements positioned on the branches to pinpoint flow observable islands. Subsequently, these islands are merged based on the available injection measurements.
It is assumed that active and reactive power measurements are paired, indicating a standard observability analysis. In this analysis, islands formed by active power measurements correspond to those formed by reactive power measurements.
Arguments
To define flow observable islands, this function necessitates the composite types PowerSystem
and Measurement
.
Returns
The function returns an abstract type Island
, containing information about the islands:
island
: a list enumerating observable islands with indices of buses;bus
: the positions of buses in relation to each island;tie
: tie data associated with buses and branches.
Examples
Find maximal islands for the provided set of measurements:
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
statusWattmeter!(system, device; inservice = 15)
device.varmeter.reactive.status = copy(device.wattmeter.active.status)
islands = islandTopological(system, device)
JuliaGrid.restorationGram!
— MethodrestorationGram!(system::PowerSystem, device::Measurement, pseudo::Measurement,
islands::Island; threshold)
Upon identifying the islands
, the function incorporates measurements from the available pseudo-measurements in the pseudo
variable into the device
variable to reinstate observability. If the abstract type Island
is derived from wattmeters, candidates for restoring observability include active power measurements and bus voltage angle measurements from the pseudo
variable. This method relies on reduced coefficient matrices and the Gram matrix.
It is important to note that the device labels in the device
and pseudo
variables must be different to enable the function to successfully incorporate measurements from pseudo
into the device
set of measurements.
Arguments
This function requires the composite types PowerSystem
and device
, which holds measurements from which the islands
variable is obtained. To restore observability, the function uses measurements from the pseudo
variable and adds a non-redundant set from it to the device
variable.
Keyword
The keyword threshold defines the zero pivot threshold value with a default value of 1e-5
. More precisely, all computed pivots less than this value will be treated as zero pivots.
Updates
The function updates the device
variable of the Measurement
composite type.
Example
Restore observability for DC state estimation:
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
statusWattmeter!(system, device; inservice = 10)
statusPmu!(system, device; inservice = 0)
pseudo = measurement("pseudomeasurement14.h5")
islands = islandTopological(system, device.wattmeter)
restorationGram!(system, device, pseudo, islands)
analysis = dcWlsStateEstimation(system, device)
solve!(system, analysis)
AC State Estimation
JuliaGrid.gaussNewton
— FunctiongaussNewton(system::PowerSystem, device::Measurement, method)
The function sets up the the Gauss-Newton method to solve the nonlinaer or AC state estimation model, where the vector of state variables is given in polar coordinates. The Gauss-Newton method throuout iterations provied WLS estimator.
Arguments
This function requires the PowerSystem
and Measurement
composite types to establish the nonlinear WLS state estimation framework.
Moreover, the presence of the method
parameter is not mandatory. To address the WLS state estimation method, users can opt to utilize factorization techniques to decompose the gain matrix, such as LU
, QR
, or LDLt
especially when the gain matrix is symmetric. Opting for the Orthogonal
method is advisable for a more robust solution in scenarios involving ill-conditioned data, particularly when substantial variations in variances are present.
If the user does not provide the method
, the default method for solving the estimation model will be LU factorization.
Updates
If the AC model has not been created, the function will automatically trigger an update of the ac
field within the PowerSystem
composite type.
Returns
The function returns an instance of the PMUStateEstimation
abstract type, which includes the following fields:
voltage
: the variable allocated to store the bus voltage magnitudes and angles;power
: the variable allocated to store the active and reactive powers;method
: the system model vectors and matrices.
Examples
Set up the AC state estimation model to be solved using the default LU factorization:
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
analysis = gaussNewton(system, device)
Set up the ACC state estimation model to be solved using the orthogonal method:
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
analysis = gaussNewton(system, device, Orthogonal)
JuliaGrid.acLavStateEstimation
— FunctionacLavStateEstimation(system::PowerSystem, device::Measurement, optimizer)
The function sets up the the LAV method to solve the nonlinaer or AC state estimation model, where the vector of state variables is given in polar coordinates.
Arguments
This function requires the PowerSystem
and Measurement
composite types to establish the LAV state estimation model. The LAV method offers increased robustness compared to WLS, ensuring unbiasedness even in the presence of various measurement errors and outliers.
Users can employ the LAV method to find an estimator by choosing one of the available optimization solvers. Typically, Ipopt.Optimizer
suffices for most scenarios.
Updates
If the AC model has not been created, the function will automatically trigger an update of the ac
field within the PowerSystem
composite type.
Returns
The function returns an instance of the PMUStateEstimation
abstract type, which includes the following fields:
voltage
: the variable allocated to store the bus voltage magnitudes and angles;power
: the variable allocated to store the active and reactive powers;method
: the optimization model.
Example
using Ipopt
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
analysis = acLavStateEstimation(system, device, Ipopt.Optimizer)
JuliaGrid.solve!
— Methodsolve!(system::PowerSystem, analysis::ACStateEstimation)
By computing the bus voltage magnitudes and angles, the function solves the AC state estimation model.
Updates
The resulting bus voltage magnitudes and angles are stored in the voltage
field of the ACStateEstimation
type.
Examples
Solving the AC state estimation model and obtaining the WLS estimator:
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
analysis = gaussNewton(system, device)
for iteration = 1:20
stopping = solve!(system, analysis)
if stopping < 1e-8
break
end
end
PMU State Estimation
JuliaGrid.pmuPlacement
— FunctionpmuPlacement(system::PowerSystem, optimizer; bridge)
The function determines the optimal placement of PMUs through integer linear programming. Specifically, it identifies the minimum set of PMU locations required for effective power system state estimation, ensuring observability with the least number of PMUs.
The function accepts a PowerSystem
composite type as input to establish the framework for finding the optimal PMU placement. If the ac
field within the PowerSystem
composite type is not yet created, the function automatically initiates an update process.
Additionally, the optimizer
argument is a crucial component for formulating and solving the optimization problem. Typically, using the GLPK or HiGHS solver is sufficient. For more detailed information, please refer to the JuMP JuMP documenatation.
Keyword
The bridge
keyword enables users to manage the bridging mechanism within the JuMP package.
Returns
The function returns an instance of the PlacementPMU
type, containing variables such as:
bus
: bus labels with indices marking the positions of PMUs at buses;from
: branch labels with indices marking the positions of PMUs at "from" bus ends;to
: branch labels with indices marking the positions of PMUs at "to" bus ends.
Note that if the conventional understanding of a PMU involves a device measuring the bus voltage phasor and all branch current phasors incident to the bus, the result is saved only in the bus variable. However, if we consider that a PMU measures individual phasors, each described with magnitude and angle, then measurements are needed at each bus in the bus
variable, and each branch with positions given according to from
and to
variables.
Example
using GLPK, Ipopt
system = powerSystem("case14.h5")
analysis = acOptimalPowerFlow(system, Ipopt.Optimizer)
solve!(system, analysis)
current!(system, analysis)
placement = pmuPlacement(system, GLPK.Optimizer)
device = measurement()
@pmu(label = "PMU ?: !")
for (bus, i) in placement.bus
Vi, θi = analysis.voltage.magnitude[i], analysis.voltage.angle[i]
addPmu!(system, device; bus = bus, magnitude = Vi, angle = θi)
end
for branch in keys(placement.from)
Iij, ψij = fromCurrent(system, analysis; label = branch)
addPmu!(system, device; from = branch, magnitude = Iij, angle = ψij)
end
for branch in keys(placement.to)
Iji, ψji = toCurrent(system, analysis; label = branch)
addPmu!(system, device; to = branch, magnitude = Iji, angle = ψji)
end
JuliaGrid.pmuWlsStateEstimation
— FunctionpmuWlsStateEstimation(system::PowerSystem, device::Measurement, method)
The function establishes the linear WLS model for state estimation with PMUs only. In this model, the vector of state variables contains bus voltage magnitudes and angles, given in rectangular coordinates.
Arguments
This function requires the PowerSystem
and Measurement
composite types to establish the WLS state estimation model.
Moreover, the presence of the method
parameter is not mandatory. To address the WLS state estimation method, users can opt to utilize factorization techniques to decompose the gain matrix, such as LU
, QR
, or LDLt
especially when the gain matrix is symmetric. Opting for the Orthogonal
method is advisable for a more robust solution in scenarios involving ill-conditioned data, particularly when substantial variations in variances are present.
If the user does not provide the method
, the default method for solving the estimation model will be LU factorization.
Updates
If the AC model has not been created, the function will automatically trigger an update of the ac
field within the PowerSystem
composite type.
Returns
The function returns an instance of the PMUStateEstimation
abstract type, which includes the following fields:
voltage
: the variable allocated to store the bus voltage magnitudes and angles;power
: the variable allocated to store the active and reactive powers;method
: the system model vectors and matrices.
Examples
Set up the PMU state estimation model to be solved using the default LU factorization:
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
analysis = pmuWlsStateEstimation(system, device)
Set up the PMU state estimation model to be solved using the orthogonal method:
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
analysis = pmuWlsStateEstimation(system, device, Orthogonal)
JuliaGrid.pmuLavStateEstimation
— FunctionpmuLavStateEstimation(system::PowerSystem, device::Measurement, optimizer)
The function establishes the LAV model for state estimation with PMUs only. In this model, the vector of state variables contains bus voltage magnitudes and angles, given in rectangular coordinates.
Arguments
This function requires the PowerSystem
and Measurement
composite types to establish the LAV state estimation model. The LAV method offers increased robustness compared to WLS, ensuring unbiasedness even in the presence of various measurement errors and outliers.
Users can employ the LAV method to find an estimator by choosing one of the available optimization solvers. Typically, Ipopt.Optimizer
suffices for most scenarios.
Updates
If the AC model has not been created, the function will automatically trigger an update of the ac
field within the PowerSystem
composite type.
Returns
The function returns an instance of the PMUStateEstimation
abstract type, which includes the following fields:
voltage
: the variable allocated to store the bus voltage magnitudes and angles;power
: the variable allocated to store the active and reactive powers;method
: the optimization model.
Example
using Ipopt
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
analysis = pmuLavStateEstimation(system, device, Ipopt.Optimizer)
JuliaGrid.solve!
— Methodsolve!(system::PowerSystem, analysis::PMUStateEstimation)
By computing the bus voltage magnitudes and angles, the function solves the PMU state estimation model.
Updates
The resulting bus voltage magnitudes and angles are stored in the voltage
field of the PMUStateEstimation
type.
Examples
Solving the PMU state estimation model using the WLS method:
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
analysis = pmuWlsStateEstimation(system, device)
solve!(system, analysis)
Solving the PMU state estimation model using the LAV method:
using Ipopt
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
analysis = pmuLavStateEstimation(system, device, Ipopt.Optimizer)
solve!(system, analysis)
DC State Estimation
JuliaGrid.dcWlsStateEstimation
— FunctiondcWlsStateEstimation(system::PowerSystem, device::Measurement, method)
The function establishes the WLS model for DC state estimation, where the vector of state variables contains only bus voltage angles.
Arguments
This function requires the PowerSystem
and Measurement
composite types to establish the WLS state estimation model.
Moreover, the presence of the method
parameter is not mandatory. To address the WLS state estimation method, users can opt to utilize factorization techniques to decompose the gain matrix, such as LU
, QR
, or LDLt
especially when the gain matrix is symmetric. Opting for the Orthogonal
method is advisable for a more robust solution in scenarios involving ill-conditioned data, particularly when substantial variations in variances are present.
If the user does not provide the method
, the default method for solving the estimation model will be LU factorization.
Updates
If the DC model was not created, the function will automatically initiate an update of the dc
field within the PowerSystem
composite type. Additionally, if the slack bus lacks an in-service generator, JuliaGrid considers it a mistake and defines a new slack bus as the first generator bus with an in-service generator in the bus type list.
Returns
The function returns an instance of the DCStateEstimation
type, which includes the following fields:
voltage
: the variable allocated to store the bus voltage angles;power
: the variable allocated to store the active powers;method
: the system model vectors and matrices.
Examples
Set up the DC state estimation model to be solved using the default LU factorization:
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
analysis = dcWlsStateEstimation(system, device)
Set up the DC state estimation model to be solved using the orthogonal method:
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
analysis = dcWlsStateEstimation(system, device, Orthogonal)
JuliaGrid.dcLavStateEstimation
— FunctiondcLavStateEstimation(system::PowerSystem, device::Measurement, optimizer)
The function establishes the LAV model for DC state estimation, where the vector of state variables contains only bus voltage angles.
Arguments
This function requires the PowerSystem
and Measurement
composite types to establish the LAV state estimation model. The LAV method offers increased robustness compared to WLS, ensuring unbiasedness even in the presence of various measurement errors and outliers.
Users can employ the LAV method to find an estimator by choosing one of the available optimization solvers. Typically, Ipopt.Optimizer
suffices for most scenarios.
Updates
If the DC model was not created, the function will automatically initiate an update of the dc
field within the PowerSystem
composite type. Additionally, if the slack bus lacks an in-service generator, JuliaGrid considers it a mistake and defines a new slack bus as the first generator bus with an in-service generator in the bus type list.
Returns
The function returns an instance of the DCStateEstimation
abstract type, which includes the following fields:
voltage
: the variable allocated to store the bus voltage angles;power
: the variable allocated to store the active powers;method
: the optimization model.
Example
using Ipopt
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
analysis = dcLavStateEstimation(system, device, Ipopt.Optimizer)
JuliaGrid.solve!
— Methodsolve!(system::PowerSystem, analysis::DCStateEstimation)
By computing the bus voltage angles, the function solves the DC state estimation model.
Updates
The resulting bus voltage angles are stored in the voltage
field of the DCStateEstimation
type.
Examples
Solving the DC state estimation model using the WLS method:
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
analysis = dcWlsStateEstimation(system, device)
solve!(system, analysis)
Solving the DC state estimation model using the LAV method:
using Ipopt
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
analysis = dcLavStateEstimation(system, device, Ipopt.Optimizer)
solve!(system, analysis)
Bad Data Analysis
JuliaGrid.residualTest!
— FunctionresidualTest!(system::PowerSystem, device::Measurement, analysis::StateEstimation;
threshold)
The function conducts bad data detection and identification using the largest normalized residual test, subsequently removing measurement outliers from the measurement set. It can be executed after obtaining WLS estimator.
Arguments
This function requires the composite types PowerSystem
and Measurement
, along with an abstract type. The abstract type StateEstimation
can have the following subtypes:
ACStateEstimation
: conducts bad data analysis within AC state estimation;PMUStateEstimation
: conducts bad data analysis within PMU state estimation;DCStateEstimation
: conducts bad data analysis within DC state estimation.
Keyword
The keyword threshold
establishes the identification threshold. If the largest normalized residual surpasses this threshold, the measurement is flagged as bad data. The default threshold value is set to threshold = 3.0
.
Updates
In case bad data is detected, the function removes measurements from the coefficient
and precision
matrices, and mean
vector within the DCStateEstimation
type. Additionally, it marks the respective measurement within the Measurement
type as out-of-service.
Returns
The function returns an instance of the BadData
type, which includes:
detect
: returnstrue
after the function's execution if bad data is detected;maxNormalizedResidual
: denotes the value of the largest normalized residual;label
: signifies the label of the bad data;index
: represents the index of the bad data.
Example
Obtaining the solution after detecting and removing bad data:
system = powerSystem("case14.h5")
device = measurement("measurement14.h5")
analysis = dcWlsStateEstimation(system, device)
solve!(system, analysis)
outlier = residualTest!(system, device, analysis; threshold = 4.0)
solve!(system, analysis)