Disclaimer: This is the authors' version of an article published in early access of IET-

Communication, Control, and Security Challeneges for Smart Grid as BOOK

CHAPTER. Changes were made to this version by the publisher prior to publication.
DOI: 10.1049/PBPO095E.

Chapter 6

Bad Data Detection in Smart Grid

Authors: Haris M. Khalid, and Ahmed Al-Durra

Affiliations: H. M. Khalid is with the department of Electrical and Electronics

Engineering, Higher Colleges of Technology (HCT), Sharjah 7947, UAE. Email:

harism.khalid@yahoo.com, Website: www.harismkhalid.com

A. Al-Durra is with the department of Electrical Engineering, The Petroleum

Institute (PI), Abu Dhabi, UAE. Email: aaldurra@pi.ac.ae

Abstract: This chapter will discuss bad data detection techniques and their

application in oscillation monitoring. Utilization of synchrophasor measurements

for wide-area monitoring applications enables system operators to acquire real-

time grid information. However, intentional injections of false synchrophasor

measurements can potentially lead to inappropriate control actions, jeopardizing

the security, and reliability of power transmission networks. An attacker can

compromise the integrity of the monitoring algorithms by hijacking a subset of

sensor measurements and sending manipulated readings. Such approach can result

to wide-area blackouts in power grids. This chapter considers bad-data detection

techniques with special focus on oscillation monitoring. To achieve an accurate

supervision, a Bayesian inference technique has been discussed for each

monitoring node using a distributed architecture.

Index Terms: Bayesian, cyber-physical systems, cyber security, data-injection

attacks, inter-area oscillations, phasor measurement unit (PMU), power system

monitoring, power system stability, real-time measurements, situational awareness,

smart grid, synchrophasor, wide area monitoring system (WAMS).

6.1 Introduction
Due to the increasing dependency of digital measurements for monitoring and

control applications, bad-data attack is an emerging threat. If a sensor is

successfully attacked, its stored information can be compromised [1-4]. Given the

criticality of power systems in the context of the national security, wide area

monitoring systems (WAMS) applications such as oscillation detection is an

attractive attack target [5]. If a PMU is successfully attacked, its stored information

can be unnoticeably compromised. This can result to significant impact on public

safety and economic losses [6-10]. Currently, it is a time-consuming task to detect

and identify data-attacks as an adversary can choose the site of attack judiciously

and design the attack vector carefully [11].

Many methods have been proposed to identify abnormal data segments and isolate

attacked sensors in recent years. Most of them are published to enhance static

application like state estimation [6, 12-13], power flow analysis [8-9, 14], and

electricity market [7]. Static monitoring applications primarily focus on monitoring

the operating point of the system and address slow dynamics in the range of

minutes to hours [15]. In contrast, few have been proposed for dynamics

monitoring applications, which tracks transient dynamics in the order of seconds or

less.
6.2 Possible Approaches
6.2.1 State Estimation
Several methodologies in the areas of state estimations were developed over the

past decades. Literature on the types of state estimation algorithms were presented

in [16-19]. State estimation also has different approaches based on application of

the algorithms such as conventional state estimation [16], distributed state

estimation or multi area state estimation [20]. Depending on the timing and

evolution of the estimates, state estimation schemes may be broadly classified into

two basic distinct paradigms: static state estimation and dynamic state estimation

[21]. Another extension of static state estimation also included the Sequential state

estimation which has the advantage of being able to perform updates with partial

measurement set [22]. This enabled the method to address the problem of data loss

and bad data. A static state estimation algorithm based on linear programming

known as least absolute value was also developed in [23, 24]. Generally, under

normal operating conditions, the power system is regarded as a quasi-static system

that changes steadily but slowly [17]. Therefore, in order to continuously monitor

the power system, state estimators must be executed at short intervals of time. But

with the inherent expansion of power systems, with the increase of generations and

loads, the system becomes extremely large for state estimation to be executed at

short intervals of time since it requires heavy computation resources. Therefore, a

technique known as tracking state estimation [25, 26] was developed. Once state

estimates were calculated, the method simply update the next instant of time using

a new measurement set obtained for that instant, instead of again running the entire

static state estimation algorithm. Tracking estimators help energy management

systems to keep track of the continuously changing power system without actually

having to execute the entire state estimation algorithm. This allows continuous

monitoring with reasonable utilization of computing resources.

6.2.2 Weighted Least Squares

One of the most commonly used types of static state estimation in utilities is the

weighted least squares methodology [18]. It was formulated as an optimization

problem with a notion of minimizing the squares of the differences between the

measured and estimated values calculated using the corresponding power flow

equations. The weighted least square uses the Newton-Raphson algorithm to obtain

the state estimates. There have been numerous findings on different variations of

weighted least square further to improve specific aspects of the algorithm. Fast

decoupled state estimator [27, 28] is an example in which voltage magnitudes and

phase angles are processed separately. The voltage magnitude values are concerned

with the reactive power measurements while angles were related to active power

measurements. Regularized least square for power systems in [29] proposed a type

of weighted least square that was able to function in cases of partial observability.

6.2.3 Dynamic State Estimation

The Extended Kalman Filter (EKF) is the most widely used algorithm to perform

dynamic state estimation [30]. Other forms of Kalman filters like unscented

Kalman filter [31], and Iterative EKF [32] were also proposed in the literature.

Other algorithms used to perform dynamic state estimations include Artificial

Neural networks (ANN) [33] and Fuzzy logic [34] which are also computationally
complex. Generally dynamic state estimations are well suited when the dynamics

of the power systems are smooth and follow the historical value. In other words,

they could fail to accurately estimate when there exists a bigger changes in

operating points.

6.2.4 Bad Data Analysis using Chi-Squares Test and

Normalized Residual Test
One of the essential functions of a state estimator is to detect bad measurements

and to identify and eliminate them accordingly [35]. Bad data analysis could be

performed during the estimation process or post-estimation. When using the

weighted least squares estimation algorithm for state estimation, detection and

identification of bad data is done after the estimation process by processing the

measurement residuals. The analysis is essentially based on the properties of the

residuals, including their expected probability distribution.

Chi-squares test for bad data detection was presented in [36, 37]. It uses the

properties of the chi-squares probability density function to compare with the

objective function of weighted least squares. Chi-squares test was able to detect

bad data but does not identify locations.

Alternatively, normalized residual test was able to detect as well as identify the

locations of occurrences [35, 36, 38]. Normalized residual test was developed

based on the statistical characteristics of the measurement residuals. Detection and

identification could also be accomplished by further processing of the residuals as

in the hypothesis testing identification methods [35, 36, 39]. Although both

methods used the residual sensitivity matrix to represent the sensitivity of the

measurement residuals to the measurement errors, hypothesis testing identification

was more complex and computationally costly due to the further processing of the

residuals. Hence, hypothesis testing identification was used to detect and identify

bad data in this thesis. However, hypothesis testing identification was observed to

exhibit some limitations as noted in [36]. The primary limitation was the inability

to track bad data if it occurred at critical locations. To resolve this issue,

utilizations of PMU measurements were proposed in recent literature.

6.3- Case Study: Oscillation Monitoring:

One mature dynamic monitoring application is oscillation detection [40-42]. Such

low-frequency dynamics were only observable by analyzing measurements from

PMUs. Today, various types of oscillation detection schemes have been installed in

many transmission utilities to monitor the inter-area oscillations within critical tie-

lines. Hence, oscillation detection is more likely to be subjected to intentional data-

injection attacks than other dynamic monitoring applications [5]. Moreover, these

data-atacks are assumed to take place in Phasor Measurement Units (PMUs)

installed in substations.

6.3.2 Consequences of an Attack

In an event of an attack, the following two negative consequences can occur due to

inaccurate monitoring and time complexity of the algorithms. 1) If the PMU data is

altered in a way that is not detectable as false dynamics by oscillation monitoring

schemes, the perceived observable state of the system will be wrong. This may

lead to improper control actions endangering the security of the system. 2) The

malicious intent might not be to hide the attack. An example is the denial of

service, where the system operator loses the observability in a critical region.
6.3.3 Distinction between a Fault and a Cyber-Attack:
Although both types of perturbations can lead to abnormal operations, the notion of

a fault and cyber-attack is distinctly different. A fault is considered as physical

events that affect the power grid behavior, where the inherent transient dynamics

are observable in neighboring substations and can be correlated in a time scale [43-

46]. In contrast, a cyber-attack decouples from the physical world [4,6,11,12,47].

Thus, the false dynamics embedded inside attacked measurements may not

correlate with other locations in time. The key is to establish a link with

neighboring metering devices, and perform correlation studies.

6.3.4 Difference from Static Monitoring Applications:

In contrast to static monitoring application, impacts of acting on incorrect

information are experienced relatively faster leading to more destabilizing issues.

The static monitoring approach of analyzing a state-space model derived from

linear differential algebraic equations is not suitable for tracking transient

dynamics like inter-area oscillations [5,6,13]. The fundamental issue is the

linearized equations are restrictive representation of the non-linear dynamic

transients caused by system perturbations. Another concern is the reaction time for

addressing transient dynamics are much less in comparison with static applications

like state estimation. Despite significant efforts have been invested in preventing

cyber-attacks for static monitoring applications [6, 7, 8, 9, 13, 47], researchers

have not fully investigated the impact of cyber-attacks for monitoring inter-area

oscillations. The reason is WAMS applications for monitoring transient dynamics

is still an emerging research where main focus to date has been towards application
development [16-18]. Hence, novel methodologies are needed to prevent cyber-

attacks in oscillation detection schemes.

6.4 Modeling an Attack in Oscillation Detection Schemes

In this section, the three critical research tasks to establish immunity towards

cyber-attacks for oscillation detection schemes are described.

6.4.1 State Space Representation of a Power Grid

In the field of real-time dynamic monitoring, especially for wide-area monitoring

system applications, the notion is to become less dependent on classic models and

adopt real-time system identification techniques. The reason is classic differential

equations are less representative of continuous random load variations, line

temperature variations, and other operational uncertainties. Although using

differential equation-based models are suitable for some steady-state or static

applications like state estimation or automatic generation control, it is not suitable

for monitoring electromechanical interactions of synchronous generators.

Therefore, system parameters are not extracted from offline predetermined power

system models. Instead, the proposed method extracts desired parameters from

PMU measurements.

A power grid prone to data-injection attacks can be modeled as a nonlinear

dynamical system. Continuous load perturbations are part of noise-induced

transitions, which can be expressed as:

(1)
where α is the constant matrix, is the nonlinear function representing the state

transition model, is the state variable, and the superscript r is the number of

monitored dynamic modes. For the case of oscillation detection, the state variable

represents the electromechanical oscillation, and r refers to the number of

oscillations in the subspace R. The process noise of the recursive scheme is

at the time t over a monitoring window of T time instances. Suppose a

power grid described by (1) is monitored by N number of PMUs. We propose a

distributed scheme that processes the estimated oscillatory dynamics of each PMU

nodes at a centralized track fusion center. The observation vector zi for extracting

electromechanical oscillations at the ith node is:

(2)

Note that the ith node may be subjected to an attack. The term and pi is the

number of measurements made by the ith PMU. The nonlinear function

representing the local observation matrix of the ith sensor is , and

refers to the observation noise.

6.4.2 Constraints of a Power Grid

From (1) and (2), a power grid can be governed by the following constraints:

(3)

where Xt, Wt, and Vt are assumed to have Gaussian probability distribution

function. Once the observation model is constructed from synchrophasor

measurements corrected from the affected location, the corresponding state

representation of electromechanical oscillations can be formulated in the frequency

domain.

6.4.3 Electromechanical Oscillation Model Formulation

According to [48], a measured noise-induced signal containing K number of

electromechanical oscillations can be modeled in the frequency domain. As a

result, (2) can be expressed as:

(4)

Ak is the complex amplitude of the kth mode, and are the corresponding

damping factor and oscillatory frequency respectively. The sampling time is

represented as Ts.

However, estimating oscillatory parameters for an accurate WAMS will require the

complete observability of the observation matrix. This is quiet challenging in the

presence of data-injection attacks. Locational awareness for each node is required,

considering the fact that installed PMUs may also malfunction during an attack.

This requires classification of the attack, followed by its characterization and

modeling.

6.4.4 Characterization of an Attack: An Example

An initial characterization about the unobservable attacks can be possibly induced

by Bayesian inference. Assume a bus of power grid is attacked as shown in Fig. 1.

Considering the Bayesian inference, the probabilities on a-prior distribution over

the oscillatory states at ith node are p( , and the observation matrix is p( .
The resultant posterior distribution over the observations can be represented by the

Bayesian inference as:

p( xti ) p( zti | xti )

p( xti | zti ) = (5)
p( zti )

To quantify the uncertainty of possible data-injection attacks, the density of the

predicted synchrophasor observations is required to be computed. This can be

obtained by averaging over the uncertainty of data-injection attacks on the

oscillatory states and the observation matrix. Let represent the predicted

synchrophasor observations at ith node, then can be presented in the form of

predictive distribution as:

p( z ipr ,t | zti ) = ∑ ∫ dH ti p( z ipr | H ti , xti , zti ) p( H ti | xti , zti ) p( xti | zti ) (6)
xti

is a hypothesis extracted from the observation signal about

the presence or absence of the attack signal.

This distribution will later assist in the development of the probability of attack

vectors.
 20

18 Fault Injection
Bus 16
16

voltage phase angle (degree)

14

12

10

0
0 10 20 30 40 50 60
time window (s)
(a)

18
Fault Injection
16 Bus 17

14
voltage phase angle (degree)

12

10 Deception Attack

0
0 10 20 30 40 50 60
time window (s)
(b)

Fig. 1 Profile of a) Bus 16 and b) Bus 17 with random fault injections

Once all the information about the covariance and estimated states are collected

from local PMU nodes, they will be treated at the distributed fusion center which is

an integral part of attack tolerant monitoring system.

6.4.5 Significance of Distributed Architecture towards

Information of Cyber-Attack
To create a cascading failure caused by lightly damped electromechanical

oscillations, the injected data can be assumed to imitate regular small-amplitude

load variations as seen in daily operations. Therefore, characterizing a plausible

attack or loss of information needs to be done. Furthermore, from a practical

viewpoint, we can assume attacked nodes are local as wide-area attacks are less
feasible from a geographical perspective. Based on these considerations, it is not

possible to identify abnormalities if a monitoring scheme utilizes one or local PMU

measurements. Conventionally, an oscillation monitoring scheme can be installed

in the PMU or local Phasor Data Concentrator (PDC). Estimated oscillatory

parameters are transmitted to the control center to optimize the communication

bandwidth. Building on top of this configuration, we intend to request each

recursive monitoring scheme to send additional information of its estimated

covariance matrix and state vector. As a result, a centralized track fusion center is

proposed. Similar to (5), the observation model of the track fusion center can

be expressed as:

(10)

Estimating oscillatory parameters in the presence of abnormal or attacked nodes

will require the complete observability of the oscillation observation matrix. This

requires the calculation of correlation information from the initial estimates of the

observation model. Treated correlation information will then assist in the removal

of faulty oscillatory parameters to ensure an attack-free update of oscillatory

parameters to the system operators. Based on this concept, a refined covariance

matrix generated is then sent back to each monitoring node to improve the

observability of the wide-area dynamics of inter-area oscillations. However, an

accurate power oscillation monitoring scheme is required to operate near real-time.

This can be very challenging in the presence of data-injection attacks. To reduce

the computational effort of determining the initial estimates and the error
covariance matrix at each PMU node, diagonalization of the system model into

subsystems can be proposed.

6.4.6 Diagonalization of a System into Subsystems:

Note the attacked system at node i can be diagonalized up to N number of

subsystems. Considering the diagonalization of N=2 subsystems, using the theory

of robust eigenvalue placement, the system (4) and (5) can be decomposed into L

and R non-singular matrices.

*
α 0 κ 0 ψ 1  i  H1i 
Lα R =  1 , Lκ R =  1 , Lψ R = , H R =  i (11)
α 2

0 κ 2 κ 3  ψ 
 2 H2 

where α1 ∈ Rn1×n2 is a non-singular lower-triangular, κ1 ∈ Rn1×n1 is quasi-lower-

triangular, κ3 ∈ Rn ×n is
2 2
non-singular lower-triangular. Transforming

xt = R[ x1,* t x2,* t ]* , where x1,t ∈ R n1 , x2,t ∈ R n2 . The system can be transformed into the

following two diagonalizable subsystems by taking the inverse of high dimensional

matrices of (1) and (2) using linear minimum variance:

x1,t +1 = κ 0 x1,t +ψ 0 wt (12)

x2,t = κ x1,t + ψ wt (13)

i i
zti = H t x1,t + v t (14)

where x1,t and x2,t are the states of subsystem 1 and subsystem 2, respectively.

κ 0 ,ψ 0 , κ ,ψ , H and v are diagonalized variables, which are computed from the

inverse of weighted matrices α1 and κ3 . Note in the subsystem transformation,

only first subsystem will have the prediction and filtering stage, whereas the rest of

N −1 subsystems will only have filtering stage. Each subsystem is a smaller matrix
than the original model, which would then improve the computing speed required

to update the covariance matrix at each monitoring instance. Once the subsystems

are constructed from the system affected by the data-injection attacks, the

interactions between them shall be evaluated. Moreover, by handling the noise and

state constraints of (3), the immunity of the estimation results during data-injection

can be increased. Referring to (12)–(14), the resultant noises wt and ν t will have

the diagonalizable expected value:

  wt  
E   1  ,  wt ν t   = Qt1,2δ t1,2
* 2*

 ν t   


where Qt1,2 is the process noise correlation factor between subsystem 1 and 2, δ t1,2

is the Kronecker delta function used for shifting the integer variable after the

presence or absence of noise.

In this step, we will integrate a centralized filter to remove estimated parameters

from attacked sensor nodes while providing accurate covariance matrix for the

individual monitoring nodes. This established a closed loop monitoring system.

Thus, the resilience of inter-area oscillation detection against data-injection attacks

can be improved.

6.4.7 Detection Bad-Data using Initial Observation Analysis:

Once the probability of attack vectors is developed, the attack can be detected by

doing an initial observation analysis of the measurements. This can be achieved by

taking the difference between the given and predicted observation of the oscillation

state:
 T
Ζit +1 = [ zti+1 − z ipr ,t +1 ] = ∑ψ t*−1θtH (1) ∆H ti + υti (15)
t =1

where the vector Ζit +1 is the innovation calculated for i-th node. zti+1 and z ipr ,t +1 are

the data-injection free (nominal) and predicted affected observation outputs,

δ zti
respectively. ∆H = ∆H − H is the perturbation in H . θ =
i i i i i
is the gradient
t d ,t t
δ H ti*t t

used to identify the perturbation due to data-injection attacks. ψ t is the data vector

formed from past outputs and reference inputs at each node.

Bibliography

[1] A. G. Tartakovsky, B. L. Rozovskii, R. B. Blažek, and H. Kim, "A novel

approach to detection of intrusions in computer networks via adaptive sequential

and batch-sequential change-point detection methods," IEEE Transactions on

Signal Processing, vol. 54, pp. 3372-3382, 2006.

[2] S. Zhu, S. Setia, S. Jajodia, and P. Ning, "An interleaved hop-by-hop

authentication scheme for filtering of injected false data in sensor networks," in

IEEE symposium on security and privacy, 2004, pp. 259-271.

[3] V. Shukla and D. Qiao, "Distinguishing data transience from false injection

in sensor networks," in 4th Annual IEEE Communications Society Conference on

Sensor, Mesh and Ad Hoc Communications and Networks, 2007, pp. 41-50.

[4] F. Pasqualetti, F. Dorfler, and F. Bullo, "Control-Theoretic Methods for

Cyberphysical Security: Geometric Principles for Optimal Cross-Layer Resilient

Control Systems," IEEE Control Systems, vol. 35, pp. 110-127, 2015.
[5] H. M. Khalid and J. C.-H. Peng, "Immunity toward data-injection attacks

using multisensor track fusion-based model prediction," IEEE Transactions on

Smart Grid, in press.

[6] F. Pasqualetti, F. Dorfler, and F. Bullo, "Attack detection and identification

in cyber-physical systems," IEEE Transactions on Automatic Control, vol. 58, pp.

2715-2729, 2013.

[7] X. Le, M. Yilin, and B. Sinopoli, "Integrity data attacks in power market

operations," IEEE Transactions on Smart Grid, vol. 2, pp. 659-666, 2011.

[8] A. H. Mohsenian-Rad and A. Leon-Garcia, "Distributed internet-based load

altering attacks against smart power grids," IEEE Transactions on Smart Grid, vol.

2, pp. 667-674, 2011.

[9] H. Yi, M. Esmalifalak, N. Huy, Z. Rong, H. Zhu, L. Husheng, et al., "Bad

data injection in smart grid: attack and defense mechanisms," IEEE

Communications Magazine, vol. 51, pp. 27-33, 2013.

[10] T. T. Kim and H. V. Poor, "Strategic protection against data injection

attacks on power grids," IEEE Transactions on Smart Grid, vol. 2, pp. 326-333,

2011.

[11] S. Cui, Z. Han, S. Kar, T. T. Kim, H. V. Poor, and A. Tajer, "Coordinated

data-injection attack and detection in the smart grid: A detailed look at enriching

detection solutions," IEEE Signal Processing Magazine, vol. 29, pp. 106-115,

2012.

[12] M. Ozay, I. Esnaola, F. Vural, S. R. Kulkarni, and H. V. Poor, "Sparse

attack construction and state estimation in the smart grid: Centralized and
distributed models," IEEE Journal on Selected Areas in Communications, vol. 31,

pp. 1306-1318, 2013.

[13] H. Fawzi, P. Tabuada, and S. Diggavi, "Secure estimation and control for

cyber-physical systems under adversarial attacks," IEEE Transactions on

Automatic Control, vol. 59, pp. 1454-1467, 2014.

[14] Y. Yuan, Z. Li, and K. Ren, "Modeling load redistribution attacks in power

systems," IEEE Transactions on Smart Grid, vol. 2, pp. 382-390, 2011.

[15] C. Rehtanz, J. Béland, G. Benmouyal, S. Boroczky, C. Candia, D. Cirio, et

al., "Wide area monitoring and control for transmission capability enhancement,"

CIGRE Technical Brochure, 2007.

[16] A. Monticelli, "Electric power system state estimation," Proceedings of the

IEEE, vol. 88, pp. 262-282, 2000.

[17] N. Shivakumar and A. Jain, "A review of power system dynamic state

estimation techniques," in Power System Technology and IEEE Power India

Conference, 2008. POWERCON 2008. Joint International Conference on, 2008,

pp. 1-6.

[18] W.-g. Li, J. Li, A. Gao, and J.-h. Yang, "Review and Research Trends on

State Estimation of Electrical Power Systems," in Power and Energy Engineering

Conference (APPEEC), 2011 Asia-Pacific, 2011, pp. 1-4.

[19] A. Leite da Silva and D. Falcao, "Bibliography on power system state

estimation (1968-1989)," Power Systems, IEEE Transactions on, vol. 5, pp. 950-

961, 1990.
[20] A. Gómez-Expósito, A. de la Villa Jaén, C. Gómez-Quiles, P. Rousseaux,

and T. Van Cutsem, "A taxonomy of multi-area state estimation methods," Electric

Power Systems Research, vol. 81, pp. 1060-1069, 2011.

[21] Y.-F. Huang, S. Werner, J. Huang, N. Kashyap, and V. Gupta, "State

estimation in electric power grids: Meeting new challenges presented by the

requirements of the future grid," Signal Processing Magazine, IEEE, vol. 29, pp.

33-43, 2012.

[22] A. Simoes-Costa and V. Quintana, "An orthogonal row processing

algorithm for power system sequential state estimation," Power Apparatus and

Systems, IEEE Transactions on, pp. 3791-3800, 1981.

[23] A. Abur and M. K. Celik, "Least absolute value state estimation with

equality and inequality constraints," Power Systems, IEEE Transactions on, vol. 8,

pp. 680-686, 1993.

[24] M. K. Celik and A. Abur, "A robust WLAV state estimator using

transformations," Power Systems, IEEE Transactions on, vol. 7, pp. 106-113,

1992.

[25] A. S. Debs and R. Larson, "A dynamic estimator for tracking the state of a

power system," Power Apparatus and Systems, IEEE Transactions on, pp. 1670-

1678, 1970.

[26] D. Falcao, P. Cooke, and A. Brameller, "Power system tracking state

estimation and bad data processing," Power Apparatus and Systems, IEEE

Transactions on, pp. 325-333, 1982.

[27] A. Garcia, A. Monticelli, and P. Abreu, "Fast decoupled state estimation

and bad data processing," Power Apparatus and Systems, IEEE Transactions on,

pp. 1645-1652, 1979.

[28] A. Monticelli, "Fast Decoupled State Estimator," in State Estimation in

Electric Power Systems, ed: Springer, 1999, pp. 313-342.

[29] M. C. de Almeida, A. V. Garcia, and E. N. Asada, "Regularized least

squares power system state estimation," Power Systems, IEEE Transactions on,

vol. 27, pp. 290-297, 2012.

[30] T. Zhai, H. Ruan, and E. E. Yaz, "Performance evaluation of extended

Kalman filter based state estimation for first order nonlinear dynamic systems," in

Decision and Control, 2003. Proceedings. 42nd IEEE Conference on, 2003, pp.

1386-1391.

[31] G. Valverde and V. Terzija, "Unscented Kalman filter for power system

dynamic state estimation," IET generation, transmission & distribution, vol. 5, pp.

29-37, 2011.

[32] M. Brown Do Coutto Filho and J. S. de Souza, "Forecasting-aided state

estimation—Part I: Panorama," Power Systems, IEEE Transactions on, vol. 24, pp.

1667-1677, 2009.

[33] A. Sinha and J. Mondal, "Dynamic state estimator using ANN based bus

load prediction," Power Systems, IEEE Transactions on, vol. 14, pp. 1219-1225,

1999.

[34] J.-M. Lin, S.-J. Huang, and K.-R. Shih, "Application of sliding surface-

enhanced fuzzy control for dynamic state estimation of a power system," Power

Systems, IEEE Transactions on, vol. 18, pp. 570-577, 2003.

[35] E. Handschin, F. Schweppe, J. Kohlas, and A. Fiechter, "Bad data analysis

for power system state estimation," Power Apparatus and Systems, IEEE

Transactions on, vol. 94, pp. 329-337, 1975.

[36] A. Abur and A. G. Exposito, Power system state estimation: theory and

implementation: CRC Press, 2004.

[37] H. M. Merrill and F. C. Schweppe, "Bad data suppression in power system

static state estimation," Power Apparatus and Systems, IEEE Transactions on, pp.

2718-2725, 1971.

[38] J. Chen and A. Abur, "Placement of PMUs to enable bad data detection in

state estimation," Power Systems, IEEE Transactions on, vol. 21, pp. 1608-1615,

2006.

[39] T. Van Cutsem, M. Ribbens-Pavella, and L. Mili, "Hypothesis testing

identification: a new method for bad data analysis in power system state

estimation," Power Apparatus and Systems, IEEE Transactions on, pp. 3239-3252,

1984.

[40] C. Lu, B. Shi, X. Wu, and H. Sun, "Advancing China's smart grid: Phasor

measurement units in a wide-area management system," IEEE Power and Energy

Magazine, vol. 13, pp. 60-71, 2015.

[41] W. Sattinger and G. Giannuzzi, "Monitoring continental europe: An

overview of WAM systems used in Italy and Switzerland," IEEE Power and

Energy Magazine, vol. 13, pp. 41-48, 2015.

[42] V. Madani, J. Giri, D. Kosterev, D. Novosel, and D. Brancaccio,

"Challenging changing landscapes: Implementing synchrophasor technology in

grid operations in the WECC region," IEEE Power and Energy Magazine, vol. 13,

pp. 18-28, 2015.

[43] Y. Zhang and J. Jiang, "Bibliographical review on reconfigurable fault-

tolerant control systems," Annual reviews in control, vol. 32, pp. 229-252, 2008.

[44] M. Blanke and J. Schröder, Diagnosis and fault-tolerant control vol. 2:

Springer, 2006.

[45] C. Hajiyev and F. Caliskan, Fault diagnosis and reconfiguration in flight

control systems vol. 2: Springer Science & Business Media, 2013.

[46] F. García-Nocetti, Reconfigurable Distributed Control: Springer Science &

Business Media, 2005.

[47] F. Skopik and P. D. Smith, Smart grid security: Innovative solutions for a

modernized grid: Elsevier Science, 2015.

[48] J. F. Hauer, "Application of Prony analysis to the determination of modal

content and equivalent models for measured power system response," IEEE

Transactions on Power Systems, vol. 6, pp. 1062-1068, 1991.

Haris M. Khalid (M'13) received his B.S. (Hons.) degree in Mechatronics and

Control Systems Engineering from University of Engineering and Technology

(UET), Lahore, Pakistan, in 2007, and the M.S. and Ph.D. degrees in Control

Systems Engineering from King Fahd University of Petroleum and Minerals

(KFUPM), Dhahran, Kingdom of Saudi Arabia, in 2009 and 2012, respectively. He

then joined Distributed Control Research Group (DCRG) at KFUPM, as a

Research Fellow. In 2013, he joined as a Research Fellow with the Power Systems

Research Laboratory (PSRL) at Department of Electrical Engineering and

Computer Science, Institute Center for Energy, Masdar Institute of Science and

Technology (MI), Masdar City, United Arab Emirates, which is a MI-MIT

Cooperative Program with Massachusetts Institute of Technology (MIT),

Cambridge, MA, USA. Currently, he is a Visiting Scholar with Petroleum Institute,

Abu Dhabi, UAE.

His research interests are power systems, cyber-physical systems, electric vehicles,

signal processing, applied mathematics, fault diagnostics, filtering, estimation,

performance monitoring, and battery management systems. He has authored 40+

peer-reviewed publications, which includes 1 Monograph, 6 IEEE Transactions, 3

IET Journals, 2 Elsevier Journals, 3 Springer Journals and 15+ peer-reviewed

International Conferences.

Ahmed Al-Durra (S'07-M'10-SM'14) received the B.S., M.S., and PhD in

Electrical and Computer Engineering from the Ohio State University in 2005,

2007, and 2010, respectively. For his M. Sc. degree, he investigated the application

of several nonlinear control techniques on automotive traction PEM fuel cell

systems. He conducted his PhD research at the Center for Automotive Research in

the Ohio State University on the applications of modern estimation and control

theories to automotive propulsion systems. At the present, he is an Associate

Professor in the Electrical Engineering Department at the Petroleum Institute, Abu

Dhabi, UAE. He obtained the PI Research & Scholarship Award for Junior Faculty

in 2014. He has one U.S. Patent Application and co-published one book titled

“Modeling and Control Aspects of Wind Power Systems.” His research interests

are application of estimation and control theory in power system stability, Micro

and Smart Grids, renewable energy, and process control. He has published over 80
scientific articles in Journals, International Conferences, and book chapters. Dr.

Ahmed has successfully accomplished several research projects at international

and national levels. He has supervised/co-supervised over 20 PhD and Master

students. He is the co-founder of Renewable Energy Laboratory at the Petroleum

Institute, and Senior Member in IEEE.