2013 46th Hawaii International Conference on System Sciences

Estimation of Transmission Line Parameters from Historical Data

Katherine R. Davis Sudipta Dutta Thomas J. Overbye Jim Gronquist
PowerWorld Corporation University of Illinois at University of Illinois at Bonneville Power
kate@powerworld.com Urbana-Champaign Urbana-Champaign Administration
dutta4@illinois.edu overbye@illinois.edu jfgronquist@bpa.gov

Abstract efforts, yet other existing sources tend to be
Power system analysis results are based on a overlooked. One resource which has not been
model of the system. The model is often assumed utilized to its full potential is the enormous volume of
known, yet it may contain errors. Model errors can historical SCADA data which is readily available to
impact the accuracy of the results. This paper many utilities.
utilizes historical data over time to extract the This work addresses the research question of
transmission line model parameters. The focus is on whether the sheer volume of available SCADA data
application of line parameter estimation techniques facilitates acceptable parameter estimates in the
for real data from a North American utility. The presence of data errors. The idea exploited is to use
approach in this paper focuses on the challenges of “data over time,” and the approach in this paper is
dealing with uncertainty and errors encountered in advantageous over methods which consider a single
the real-world data. Future applications are also snapshot of the system. The parameters of a pi-
discussed. model of a medium length transmission line are
determined from multiple measurements of complex
power and voltage magnitude at both ends of the
lines.
1. Introduction The organization of the paper is as follows. A
brief background of the estimation problem is
Model development and validation is an area of provided in Section 2. The transmission line model
growing importance for power systems. The reason used in the work and an overview of the parameter
for placing high emphasis on model accuracy is that estimation method is presented in Section 3. The
in any discipline, results are only as good as the proposed methodology is described in Section 4.
models upon which the analysis is based. Thus, Results obtained by applying the proposed
identifying model errors is an important problem. methodology on a simulated test case are presented in
This work incorporates large volumes of Section 5 and results obtained on real SCADA data
historical data collected over time to estimate the from a North American utility are presented in
values of transmission line model parameters. Real Section 6. Section 7 presents discussions on the
data often contain significant errors which do not results. Section 8 discusses future applications for
have known probability distributions. Meters used transformer model parameter estimation. Finally,
for measuring data are a common source error. A conclusions are made in Section 9.
key aspect of this work involves being able to
perform analysis such that any errors in the available
data do not corrupt the estimate. 2. Estimation Background
With respect to data collection, phasor
measurement units (PMUs) [1] have received high Many applications rely on correct underlying
visibility. Increasing PMU deployment leads to an model information. In power systems, the area where
increase in the amount of data available and a desire estimation tends to receive the most attention has
to effectively manage, send, and process the data. It been in power system state estimation [2], [3].
remains a challenge to intelligently analyze the data Parameter identification has often been studied in the
and obtain useful information to advance the state-of- context of state estimation literature as state
the-art for power system operations and control. estimators also serve as data filters. Uncertainty in
Data from PMUs is often the focus of data mining data with respect to state estimation is analyzed in
[4].

1530-1605/12 $26.00 © 2012 IEEE 2149
2151
DOI 10.1109/HICSS.2013.206

and I2 denote the magnitudes of the P1  V12G11  VV G12 cos 12 .[6]. θ2. typically. Also. which can be exactly available data. This is a In the pi model. and in time. x. shown in Figure 1. Pi Line Model The line model considered is the pi model for transmission lines. If it is possible to know the angle Historical SCADA data. most may be unavailable.1. and V2. |S2| are also known. I1. Background on the statistical analysis SCADA data. and Q2 are of data [10] provides insight into the quality of the measurements. of approximately 50 to 150 the transmission line model is Y=G+jB: miles. the current magnitudes. power factor angles are provides a fundamental expectation of what can (and also known at both line ends. these true raw measurements a means for updating the estimates online. This is the ‘raw’ SCADA data. Q1. The equations for real and reactive power flow at both ends of the line are given by the following: Figure 1. This From the current magnitudes. V2. Thus. the complex powers S1. The goal is to use available existing approaches consider only a single snapshot SCADA measurements to estimate values of r. Thus. the transmission line model is I i  Vi / Si (1) described and analyzed in the context of the measured data which is used in this work. Transmission Line Model and and voltage magnitudes. 3. results for small simulated b in Figure 1. difference across the line and the line parameters are This paper circumvents both previous limitations. Several properties of the data are important to the quality of quantities can always be derived directly from the our estimate. The assumption is that the angle systems are presented as opposed to real systems. Basic SCADA Data Properties data collected over a number of different operating points can be applied to construct a reduced The pi model equations are now considering the equivalent model at the PMU buses. the apparent powers |S1|. let b/2 denote the value of the standard model which is often used to model lines of shunt modeled at each bus. unknown. S2 at the line obtained estimate. Transmission system parameter inaccuracies can Q2. however.[5]. Pi Line Model Let V1.[8]. computed from the SCADA data. In this section. From the apparent power 3.[7]. are the line parameters and the angle difference across the line.2. recorded every five minutes difference across the line (as with PMUs). The statistical assumed known or measured quantities. Notably cannot) be estimated about the model based on the lacking in these quantities. the over several months. recorded have adverse effects on state estimation. is available and is presented in problem is considerably simplified. Section 6 of this paper. The admittance matrix of medium length [11]. Note that what is physically existing work is aimed at identifying and correcting measured by potential transformers (PTs) and current transmission model parameter errors using telemetry transformers (CTs) is only voltage and current. Transmission Line Model 3. θ1. where PMU 3.  1 1  I1 I2  r  jx  jb / 2  r  jx  r+jx Y  (2) P1 + jQ1 P2 + jQ2  1 1    r  jx r  jx  jb / 2  V1 V2 jb/2 jb/2 The real and reactive components of Y gives matrices G and B with elements Gij and Bij. every five minutes.pf. Since P1. Available Data I2. P2. A close analogy to the key concept of this work is introduced in [9]. I1. can be directly computed.pf.3. data [3]. While a few methods provide respectively. ends are directly known.

 B12 sin 12 .

The assumption is that the P2  V22G22  V2V1  G21 cos 21 . (3) 1 2  phasor quantities.

 B21 sin 21 .

 (4) measurements for this application are P1. V1. Q1. P2. 2152 2150 .

G12 sin 12 .

 B12 cos 12 .

 (5) Q1  V12 B11  VV SCADA estimation problem. some approximation is 1 2  necessary. The choice of the approximation is Q2  V22 B22  V2V1  G21 sin 21 .

 B21 cos 21 .

The Plosses  V12  V22 . (6) therefore important. the equations for the real and reactive 3.4. Estimation Equations power losses in the line are thus given by P1+P2 and Q1+Q2 respectively: The equations used to perform the SCADA estimation are presented in this section. From (2)-(6).

G11  2VV G12 cos 12 .

 (7) 1 2  following two equations for real power losses and reactive power losses contain the three unknowns (r. Qlosses  V12  V22 .

B11  2VV B12 cos 12 .

Thus. (8) 1 2  x. four out of the six equations in (3)-(8) are Plosses  I 2 r (10) independent. and b) which are time independent. The loss equations are obviously not Qlosses  V 1 2 b / 2.

 V2 b / 2.

whereas the angle difference θ12 is a function of equations is obtained by appending the contributions the operating point and changes over time. a system of time. from the measurable quantities. Future work may investigate the either the term sin(θ12) or cos(θ12). For example. and r to minimize the difference to the estimation equation for r. Thus. The unknowns Gij. As an The difficulty is that the unknowns appear in the approximation. The key difficulty is that from one snapshot. B12. b/2. Due to current This looks similar to a traditional state estimation injections from the shunt elements. Solution requires impact of the model approximation on the parameter a nonlinear iterative approach such as Newton’s estimation. Thus. the of repeated measurements to the matrix A. between the measurements and the calculated quantities. and θ12 = (θ1-θ2). it follows This avoids the problem of unknowns which depend that use of a linear technique to estimate these on time. To see Other variations of the estimation equations are this. Essentially. magnitude I in (10) and (11) is not known exactly there are four equations and four unknowns. observe that what appears in the equations is possible depending upon which approximations are always a product term of Gij and Bij multiplied with desirable to make. Based on (12) and (13). The goal of this work is Qlosses    1 b  I x 2 (13)  2  to improve the ability to directly estimate b. a bias term u can be added method to find b. This is true at least over some window of measurements. equations in a nonlinear manner. Bij. the average of I1 and I2 is taken for I. and r using data collected over multiple points in time. I differs from I1 problem except with different unknown variables. Thus. and b/2 are not dependent which represents a Ploss static error in the on time. x. the The need for approximation arises since the current unknowns are G12. there are Plosses  u  I 2 r (12) more unknowns than equations so traditional state  V  V2  2 2 estimation is not sufficient.  I 2 2 x (11) independent of the first four since they are each the sum of two other equations. In equations (3)-(6). this is state estimation. quantities will not yield complete success. and I2 which are known from (1). x. This may terms in (3)-(6) which depend on time are the be expressed as (14) or (15): following: 1 I 2 0 0  u   P  G12 cos 12 .

. B12 sin 12 .

   r   losses  G12 sin 12 .

. B12 cos 12 .

(9) 0 0 I 2 1 2 .

q (15) snapshots in time to estimate the transmission line parameters is only successful if there are fewer unknowns which depend on time than the number of As additional rows are added corresponding to linearly independent equations. this becomes an four linearly independent equations. In this case. Each additional overdetermined problem.   x    Qgen  (14) V 2  V 2 / 2     b    The proposed approach of considering multiple Aey  fp. Least squares estimation time point will produce four more equations but will minimizes the residual. in the 2153 2151 . Thus. there are additional time points. also produce four more unknowns.

MeetsFilter then begin Input: DataSet. Value filters. is equal to zero. the estimate presented. in this initial fit. nR keepIndex. Q. which is applied as a component of If row = 0 then begin Algorithm 2. nR End Foreach row of DataSet do begin End [F. allow Algorithm 1 to return a dataset Algorithm 2.add(row. End when applied. such as a<x<b. chunkSize. However. For a specified line. Regression for particular conditions.1. However. Estimation and Outlier Rejection power losses Plosses and the negative of reactive power losses Qgen. σci] = normfit(r) 2154 2152 . At each iteration. Algorithm 1. A data [DataSet. It is possible to estimate Plosses Once an initial data set is selected. as outlined in Algorithm 2. Q. r  Aey  fp. r. where ey = [u. DataSet initial screening and filtering of the raw SCADA ey = LeastSquaresEstimate(DataSet) data. σ. for selecting and returning a set of data on which the nR. analysis is to be performed. where x is any of the measured or computed variables.q (16) DataSet = DataSet (keepIndex) and results in the following estimate (14). with residuals which exceed a predetermined threshold and reject them. b]. and Function PerformRegression Input: DataSet V at both line ends. V at line ends keepIndex = (rand) to (rand+chunkSize) an impact on the quality of the initial fit. nR) point or a row refers to one set of SCADA ey = LeastSquaresEstimate(DataSet) measurements from one point in time. Then.q contains real 4. The procedure stops when the number of The procedure outlined in Algorithm 1 is responsible rows removed from the dataset during any iteration. This function also provides Output: ey = [u. r. By removing An overview of the proposed methodology is now outliers or problematic data points. Currently. the dataset is obtained from least squares. While nR > 0 then begin nR = 0 Additional screening may also be performed. Algorithm 1 returns a random ey  [AT A]1 ATfp. nR] = OutlierFilter (DataSet. Thus. the sample dataset consists of measurements of P. Procedure for Dataset Selection In summary. r. regression is and Qgen independently since the equations are applied.q (17) consecutive data chunk of size chunkSize which has undergone initial screening and value filtering. x. The vector fp. any Input: RawData. as evident in (14). b] denotes a vector of the estimated quantities. all of the data are used 4. other outlier filtering algorithms can End easily be substituted in its place. filter. regression is indicate a poor match of the particular data point to applied to detect and reject possibly wrong data the model. all data is screened simply by r = Aey-b removing data points which have all values of zero. µci. The strategy is to identify data points based on analysis of a fitted model. x. DataSet = RawData(keepIndex) Outlier filtering is performed as indicated in Foreach row of DataSet do begin Algorithm 3. High residuals rejection. a data selection module returns improves. Initially. While Algorithm 3 is the currently Remove row End implemented. whichLine.2. the residuals or the dataset and also accomplishes data ‘screening’ deviations are computed using (16) with the current functions such as value filtering and initial bad data values of the estimated parameters. errors which exist in the initial dataset can still have Output: DataSet consisting of P. The regression algorithm thus progressively identifies and removes bad data and computes a new Function DataSelection estimate from the refined data set. An initial fit to completely decoupled.Index). Dataset Selection data points have residuals within the specified threshold. outliers can have a significant impact on the estimate. for a given set of data. clear keepIndex Foreach row of DataSet do begin Function OutlierFilter If row. The process continues and iteratively fits the data to the refined data set until all 4. Output: DataSet. xi] = ksdensity(r) [µ. Proposed Estimation Approach indiscriminately. r.

02 x: 0. 0. Estimates on Table 1.5 0.99e-5 0. Then. 200 perturbations are made of each point. Matlab functions ksdensity and normfit units of P. If the error residual for r: 0.01 pu.002 Data reference. estimates of mean µ and standard deviation σ of the data with a 95% confidence interval.005 0.005 0 -0. The per unit base is 100 MVA. distribution estimate for the sample vector r. The objective of Algorithm 3 is to detect and (3. The system used for the simulations is the seven bus system shown in Figure 2. Application to real data is presented in the next 0. Seven-bus error-free data Qgen plot 5. with If residual(row) > threshold then begin model parameters: r=0.9901e-005 0. This reflects the fact that during each quasi- Figure 2.02 pu. MVAr have been used from the Statistics toolbox.020058 0.006 5.2. respectively. the point is u: -1.010055 0. and V measurements are MW.01 residual from the zero mean.020058 remove ‘bad’ data characterized by deviation of 0. Application to Simulated Data 0.9 1 A program allows SimAuto to retrieve SCADA data Current(pu) from the simulated system.4 0.005 Data Estimate Model -0. 0 0. By repeating this procedure. PowerWorld 0. x=0.8 0. The values of the estimates are given in function returns a vector of density values F Table 2.008 removed.030161 0. and the model are shown in Figure normfit(r) fits r to a normal distribution and returns the 3 and Figure 4. b=0. σ) The estimated line is between Bus 3 and Bus 4.003 simulated data from a test system to serve as a 0. the procedures are applied to 0. Q.2 0. This as expected. Ploss(pu) 0.01 Qgen(pu) section. The function data.6 0. Simulations provide a reference case since b: 0.4 0. The and per unit. Using only these points function ksdensity (r) is used to compute a probability with no added error.2 0. the estimate matches the model. The In Algorithm 3.7 0.01 0. 2155 2153 . the system is Figure 3.03 pu. Seven bus study system steady-state operating point. 4) -1.010055 0.001 Estimate Model Simulator is used via SimAuto and script commands. Outlier Filtering A simulated SCADA dataset for line (3.5 0.030161 obtained.009 a data point exceeds the threshold.3 0.004 In this section. To perform these simulations.7 0. the SCADA values do not change much but may change slightly. Study system noise-free estimates the bounds of the confidence intervals are given by µci Line u (pu) r (pu) x (pu) b (pu) and σci.6 0.9 1 Current(pu) Figure 4. Simulated Error-Free Data End End Algorithm 3. Simulated Data with Error From the data for the five ‘error-free’ operating conditions above. the estimate. threshold = f(r.8 0.007 0.3 0.4) is obtained from five different operating points. Remove row from DataSet Increment(nR) End 5. Thus.1. Seven-bus error-free data Ploss plot perturbed to a new operating point. data for a number of operating points is 0. Real and reactive power plots of the actual evaluated at the points in vector xi.015 both the model and the data are exactly known.

j) in Table 3 denotes a line α (3.1 1. It is interesting to note that the five operating Table 3. tabulated in Table 3. Existing Model Values for Lines points result in five distinct clusters of points on the Line r (pu) x (pu) b (pu) graphs.2 able to present these results for actual SCADA data.00999 0.05 system. The issue is that when normal probability distribution of mean zero and simulating data.9079 presence of these bands.02 0 6.1 using real world data can be extremely challenging as x: 0. u: 0.01054 0. However.2 0.06 three transmission lines illustrated in Figure 7.2.02931 0. Thus. Qgen(pu) 0. 0.0012165 Model 0. These 1000 points are used to compute the how one should perturb the system to obtain estimate.5 0. 4) 0.00056 0.017988 Each line identifier α=(i. and the corresponding possible ways to change the system and obtain a new estimates are shown in Table 2. the authors have observed the A=(1.75194 small set of points is examined.8 0. Results for these Figure 6.3 0.04 Gaussian or known probability distribution. parameter estimation 0.00025 0.32042 measurements of the same operating point seem to cause this feature.9 1 1. 4) 0. Each some sets of operating points facilitate better measurement is per-unitized and randomized by a estimates than others.08 0. Seven-bus noisy dataset Qgen plot three lines are presented. = 1000 hypothetical measurement data points.006369 0. multiple noisy C=(3. it is extremely valuable to be 0.05 0 The proposed procedures are applied to historical SCADA data from a real North American power -0.08 b: 0. and 0.017988 Estimate evident in the results and discussions presented in the Model 0. Study System . 2) 0.7 Current(pu) 0. the most value is obtained from testing the Estimate procedure on real data. Comparison of Model to Data 2156 2154 . Identifiers of buses and lines in the real system have been made anonymous.00445 0.010153 0. From a -0.0063688 Data studied. The -0.1 6.02932 0.020321 between bus i and bus j.1 practical point of view.8 0. Bus 1 Bus 2 Bus 3 Bus 4 Table 2. As indicated by the estimates for (3. The results from the new dataset are shown representative operating points since there are infinite in Figure 5 and Figure 6. 3) 0.1 1. Line A Line B Line C the accuracy of the r estimate is considerably improved when the static error u is not estimated. SCADA data collected over several months is available for the -0. 4)* . 4)* in Table 2.1. Study system estimates with noise Figure 7.4 0.from the five exact operating points. In real data. operating point.4 0.7 0.02 per unit with a realistic distribution. especially when a relatively B=(2.6 0. 6. 2 MVAr.North American Utility -0. Thus. Three Lines with Available Data Line u (pu) r (pu) x (pu) b (pu) (3. we create 200*5 As may be evident when comparing the results.6 0. rather than speculate on what representative operating points exist that should be 0.029319 Data 0.9 1 1. Figure 5.02 -0.5 0.15 r: 0.06 following sections. Application to Real SCADA Data Ploss(pu) 0.3 0. it is challenging to reproduce data standard deviation 2 MW.2 0. Real world data rarely follows a 0. from Figure 7.2 model values of the line parameters for these lines are Current(pu) also available. Seven-bus noisy dataset Ploss plot Estimation on artificially constructed data is a special/trivial case. The question arises of voltage.00052 0.04 From a North American utility.0012165 0.

1 0.8 0 -0.0067853 each line. and C lead to reasonable and consistent parameter Figure 10. 14. 0. Figures 1.000 data points 2 Data is selected according to Algorithm 2.4 0.5 0.1 Model 0.6 0.1. Line B plots.2 -0.9 estimated from the reactive power losses. 0. 0.90628 Figure 11.2.5 0.1 -0.1 0.3 Current(pu) 0.039917 estimates for all of the lines are given in Table 4.05 0 0. Ploss vs. and 18 plot Ploss and Qgen vs.2 Data 1 Estimate 0.5 0.15 -0. 15.8 Std: 0.5 (red dotted lines).2 -0.01163 0 b: 0.2 0 6. Qgen residual distribution (Line A) 2157 2155 . I for r: 0.4 0 2 4 6 8 10 12 Data Current(pu) 1 Estimate Model Figure 8. Line A plots.2 0.5 estimated and model parameter values.25 -0. while x and b are 0. Figures 9.9 Density Distribution Estimate -1 0.6 Std: 0.010794 Ploss Deviation (pu) Qgen(pu) 0. 10.3 g 1.15 0. In the next sub-sections.5 6.05 0. No value Estimate Model filtering is applied. -0. loss An initial dataset with a size of 30.5 x: 0.05 0 0. Qgen vs.8 -1.2 x: 0. Qgen vs.1 0 -0. 0 2 4 6 8 10 12 Current(pu) B.3 -0.4 -3 8 0 2 4 6 8 10 12 14 16 18 0. The red and blue lines on these plots show respectively the expected data points when using the 0.2 g 1. 13.0811e-006 0. 12.000 data points.2.15 Qgen Deviation (pu) Figure 9. I (Line A) 0.00086639 Ploss(pu) 1 u: 0. The available datasets for lines A. 16.7144 1 Qgen(pu) -0.15 -0. 10.7 Each estimate is based on 30.1 -0.2 Figure 12.5 8.05 0. and 19 show distributions of the residuals 0 about the parameter estimate as well as Gaussian distributions of the same mean and standard deviation -0.6 b: 0. Ploss residual distribution (Line A) 0.014912 -2 -2.7 Mean: 2.5 -0.551e-017 Density Distribution Estimate 0.2. These Mean: -3. 17.5 0. I (Line A) estimates. I (Line B) 0.4 0. Resistance r and static power offset u are estimated 1 from the real power losses.

303e-005 0.4 0.2 0. I (Line C) 0.5 0.25 -0.5 0. I (Line C) 1 6.5 r: 0. Qgen residual distribution (Line C) 1.1 0.6 -0.1 0.4 0.2 -0.6 0.1 0. Density Distribution Estimate 0.35742 Density Distribution Estimate Model 0.3784e-005 Density Distribution Estimate 0.039822 Ploss(pu) 0.8 0 0.1 0.25 Qgen Deviation (pu) Current(pu) Figure 13.2 0.4 -0.7 Std: 0.2 -0.5721e-017 0.15 Current(pu) Qgen Deviation (pu) Figure 14.3 0 0.05 0.1 -0.2 0.4 0.1 0.8 b: 0.05 0 0.6 0.8 Mean: -5.3 0.6 Ploss(pu) 0.1 0. Ploss residual distribution (Line B) Figure 18.1 0 0.15 0.019167 Model 0. I (Line B) Figure 17.15 0.05 0 0.00047885 0.05 0.2 u: -0.3 -0.2 0 0 2 4 6 8 10 12 14 16 18 -0.5 Ploss Deviation (pu) Current(pu) Figure 15.0086491 0.05 0 0. 1 1 x: 0.3 0.3 0 0. Qgen vs.9 Estimate 0.2 0.8 Std: 0.2 0.15 -0.0044154 Data 0. Qgen residual distribution (Line B) Figure 16.3 0.2 -0.15 -0.2.1 0 -0.1 Std: 0.035339 0.2 -0.3 0.1 -0.065345 0.2 -0.00025763 Data Density Distribution Estimate 0.05 0.7 1 Data 0.6 Qgen(pu) 0.6 Estimate Std: 0.1 -0.4 u: -0.7 0.4 0 4 0 2 4 6 8 10 12 14 -0. Ploss residual distribution (Line C) 2158 2156 .1 -0.2 Model 1 Mean: 2.4 0 0 2 4 6 8 10 12 14 -0.7 0.7123e-017 0.5 0.15 -0.1 -0.4 0. Ploss vs.4 0.9 Mean: -1.25 Ploss Deviation (pu) Figure 19.3 r: -0.3.4 0.010291 Estimate 1.2 0. Line C plots.2 -0.1 0. Ploss vs.2 0.9 Mean: 1.8 0.2 -0.25 -0.2 -0.

000305 pu. Estimates from 30. estimation is done after applying regression transformers. I (Line C.0077929 0. as evident from the plots for distribution in per unit. there is no way to validate the final observed to vary linearly with terminal voltage [14].4 0 2 4 6 8 10 12 14 Current(pu) 7. etc. as the dataset which is chosen ultimately impacts the quality of the final estimate. 2) 0. The data for reactance and shunt estimation for the where K is the constant to be estimated and is three lines was ‘better’ than the data for resistance specific to the transformer. transmission lines are not known or rather cannot be GIC-related transformer reactive power losses are exactly known. In the simple of residuals. When we use multiple months of SCADA SCADA data.2 The estimates of the line parameters based on the 0. However.1 data above are given in Table 4 below. The estimated resistance for Line A is A=(1. IGIC is given by contain systemic biases which are not accounted for in the approach.004401 0.71445 due to the fact that the available model resistance C=(3.00030975 0. rise in temperature [12]. Vpu is the terminal voltage estimation. the estimated Table 5. 2) 0.011749 0. 2159 2157 . and a corrected dataset is returned after applying this filtering in A future use of historical data to estimate model Algorithm 1.0004928 0. 3) -0. Ploss(pu) 0 Table 4.3 0. is considerably improved. I plot. results.4% increase in resistance per °C is much longer in comparison to the Lines B and C. GIC Model Validation Applications are corrected. the values provided in Table 5 are and may be represented. Measurements may also autotransformer.000257 0. any changes in the actual estimated resistance is close to the model resistance parameters are ignored. Units of K are MVAr/A. This may indicate the presence of measurement sign errors. 4) 0.010794 0. Figure 20 modeling. 4) -0. The new estimate of the resistance. It is clear that ‘screening’ the data is a critical stage of the Qloss  Iline 2  xm  Vpu KIGIC (18) estimation process.7144 r:0.6. for example.0004789 0. The suspected sign errors 8.35742 u:0. Parameter Estimates for All Lines 0. However.90586 slightly higher than its model parameter.2. The latter data is noisy and the Ploss case of a wye-delta step-up transformer.000 Data Points -0. In seasonal planning models. and IGIC is the DC GIC. Since the actual parameters of the effects in large power systems is presented in [13]. available. the magnitude of the data to obtain an estimate.010256 0.010774 0.0058326 0. The estimated resistance of Line C is negative the r values are not adjusted since x values typically when the algorithms are applied on the available raw dominate.3589 value was computed using the dimensions of the line.0008664 0. 3) -0. Estimates after Regression parameters are close to the available model Line u (pu) r (pu) x (pu) b (pu) parameters.4. A more thorough coverage of the at I H  I L I GIC  (19) detection of data problems is beyond the scope of this at paper. The reason this does not actual line resistance varies with temperature.004415 0. as expected to be the best parameter estimates.1 Line u (pu) r (pu) x (pu) b (pu) A=(1. GMDs induce geomagnetically induced shows the new Ploss vs. IGIC is the values are small compared to the value of the neutral current IN. For an parameter being estimated. There occur for the other two lines may be because Line A is approximately a 0.0094121 Model -0. material resistance. Background on the problem and its for all the lines. r parameters is in geomagnetic disturbance (GMD) = 0. currents (GICs) in transmission lines and Next. Ploss vs.008649 0.3 Estimate C=(3. corrected data) As can be seen from the above plots. Discussion of Results Figure 20.010291 -0. this may be B=(2.0009012 0.90628 -0.00030502 Data -0.0067853 0.2 B=(2. for different conditions than An effect which has not been included is how actually seen in operation.01163 0.

Transmission and When the K values of transformers are not known. Rogers." IEEE Transactions on Power Apparatus and Systems. vol. Nov. D. Sept. 2. PAS-98. 5. G. J. G. Power. Acknowledgments [13] “2012 Special Reliability Assessment Interim Report: Effects of Geomagnetic Disturbances on the Bulk Power System. March 2005. vol. Al-Othman. Prentice Hall. References PAS-100. Conference and Exposition (PSCE). 1-8. as shown in this paper. 152. Apr 1993. no. 6. [7] J.594-607. 1999. 586. G.” IEEE Power Systems challenges which are not encountered in simulation. In summary. American utility. [8] J. Sasson. "Load-flow studies in the presence of Solutions (CERTS). 2. March 2012. 2004. Zhu. PAS-93. J.2. pp. [15] V. pp. Liu. Irving. Proceedings. L. 6. 2010. 10-15. vol.” submitted to IEEE Transactions on Power power systems. 1260- 1268. the Power Systems Engineering power consumption from GIC saturated Research Center (PSERC). T. Zhu. G.239. 1.” NERC. [2] A. no. pp. Overbye." IEEE Transactions on Power Systems. M. Smolleck. R. 5. Bergen. Dong. and Winter Meeting. 2nd ed. The approach is [10] J. 9. Results for both [9] K. Mineola. 2160 2158 . no. The authors would like to acknowledge support from [14] X. they are often approximated in the analysis by default [5] G. pp. al. M. vol. “Influence of system characteristics on the amplitudes of geomagnetically induced [1] A.445 vol. 233. “Real-time external equivalent for on-line contingency analysis. Irisarri. Feb. pp." IEEE Transactions on Power Systems. Kusic. M. it is preferable to use actual values transmission line parameters from SCADA to improve the meaningfulness of the GIC results. 179-190.from [15] [16] where the two DC coil currents IH and [4] A. Mandel. State estimation in electric power systems: a generalized approach. 2. “Guidelines for the selection planners to find and correct model errors based on and operation of bare ACSR conductors with regard data which is already collected. A. It is clear that “Identification of power system topology from estimation based on real measurement data presents synchrophasor data. 2153-2171. N. 2012. vol.D. Spadoni. "Improvements in network parameter error identification via synchronized phasors. 2000. Kluwer Academic Publishers. vol. Monticelli.” The value of using data collected over time is IEEE Transactions on Power Systems. needed for GIC analysis.. Phadke. Trustworthy Cyber transformers.” Electric Power Systems Research." IEEE Power Engineering Society Infrastructure for the Power Grid (TCIPG). can help validate transformer parameters which are 21. no.. [3] A. vol. V. Vittal. 11. A. et. Mohan. G. pp. parameters. 1974. 10. Power System Analysis. the proposed [11] R. to current carrying capacity. (Reprint) Dover. "Comparative the Stanford University Global Climate and Energy analysis of exciting current harmonics and reactive Project (GCEP). F." IEEE Transactions on Power Apparatus and Systems. Oct. pp. 1984. Feb 2001. The Statistical Analysis of Experimental validated on actual SCADA data from a real North Data. Parameter estimation errors. Debs. Abur." IEEE PES Power Systems Conference and The NERC report [13] identifies as an action item the Exposition. L. pp. J. Albertson.592. Consortium for Electric Reliability Technology Skarbakka. D. simulated and real data are discussed. pp." IEE ratio. Conclusion 25. Kappenman. N. May 2006. methodology is a valuable tool which can allow [12] S. 1979. pp. apparent when estimating transmission line no. 2004. geomagnetically-induced currents. A. S. 440. pp. "Identification of network parameter reactive power loss modeling. Zheng. development of tools for GIC flow and subsequent [6] J. Y. to real data is thus valuable. Feb. vol. 1982. 44-50. 1981. A. "Uncertainty IL may be measured.Generation. no. 1. "Synchronized phasor measurements in currents. Feb. "Measurement of values. J. Garrison. "Estimation of steady-state power system model parameters. [16] K." IEEE Computer Applications in Delivery. Abur. 318-322. Obviously. data. and at is the transformer turns modelling in power system state estimation. March The presentation of the proposed solution approach 2011. Sims. R. K. Dopazo. no. Distribution.Y. Kappenman.