You are on page 1of 6

State Estimation Techniques for Electric Power Distribution Systems

Barry Hayes and Milan Prodanovi´c
IMDEA Energy Institute, Madrid, Spain.

Abstract—This paper provides a survey of techniques for quality of available measurement data), many of the methods
state estimation in electric power distribution systems. While and approaches developed for “conventional” transmission-
state estimation has been applied in the monitoring and control level SE cannot be applied directly to DSSE. Hence, a
of electricity transmission systems for several decades, it has
not been widely implemented in distribution grids to date. number of SEs specifically designed for application at the
However, with the recent drive towards more actively-managed, distribution level have been proposed in the literature in
intelligent power distribution networks (“smart grids”) and recent years. However, despite the growing importance of
the improvements in monitoring and communications infras- DSSE, the authors were unable to find a relevant survey
tructure, Distribution System State Estimation (DSSE) has paper in the literature, summarising the current state of the
been receiving significant research interest. DSSE presents a
number of unique challenges due to the characteristics of art, and discussing research trends and future directions in
distribution grids, and many of the well-established methods the area of DSSE (one conference paper was found [2], but
used in transmission systems cannot be applied directly. This the literature survey in this paper is not comprehensive, and
paper provides a detailed survey of the available methods for focuses mainly on Chinese-language publications). While
DSSE, reviewing around 70 papers from the major journals. there have been several survey papers and books with liter-
In addition, it discusses the potential for applying Advanced
Metering Infrastructure (AMI) data and computational intel- ature reviews in the general field of power systems SE [3]–
ligence methods in DSSE. [11], these deal primarily with techniques and methods
applied to transmission systems, and there are none which
Keywords–state estimation; distribution networks; distributed
energy resources; smart grids focus specifically on the developments and applications of
DSSE. This paper aims to fill this gap by providing a survey
I. I NTRODUCTION of the most important techniques and algorithms currently
Since the initial development of the concept in the early available for DSSE.
1970’s [1], power system State Estimation (SE) has become This paper will also discuss the application of Advanced
a critical part of the operation and management of trans- Metering Infrastructure (AMI) data, such as smart meter
mission systems worldwide. Until recently, the application measurements, as inputs to the DSSE algorithms. Addition-
of SE at the distribution level, i.e. Distribution System State ally, the use of novel computational intelligence methods and
Estimation (DSSE), has not been of significant interest. This machine learning approaches and their potential benefits in
is largely because distribution networks have traditionally this context of DSSE will be explored. The paper is struc-
been designed and operated as passive systems, where tured as follows: Section II describes the main techniques
power flows are unidirectional and relatively easy to predict and applications of DSSE, Section III outlines the current
and manage. However, distribution networks are seeing state of the art, and highlights some of the most advanced
increasing penetrations of distributed energy resources, such methods currently available. Section IV discusses the use of
as small to medium-sized Distributed Generation (DG), AMI data and computational intelligence methods in DSSE.
demand-responsive loads, electric vehicles and devices with Finally, conclusions are drawn in Section V.
storage capability. This has led to a requirement for im- II. DSSE T ECHNIQUES AND A PPLICATIONS
proved observability in distribution systems, and the need A. Conventional Power System State Estimation
for Distribution System Operators (DSOs) to take a more
active role in monitoring and controlling the operation of the SE is used to improve system observability, check for
networks. DSSE has a crucial importance in this context. and detect errors in both system measurements and net-
Since distribution networks have different characteristics work parameters, and to mitigate against measurement and
to transmission networks (e.g. radial construction, high R/X communication system noise. Detailed summaries of the
ratios, phase imbalances, and a much lower quantity and main techniques and applications of conventional power
systems SE can be found in [3]–[5]. Fig. 1 shows a graphical
The authors kindly acknowledge the support of the European Commis- overview of the main processes and information flows. First,
sion projects on Marie Sklodowska-Curie researcher mobility action (FP7- a topology processor verifies that the network parameters
PEOPLE-2013-COFUND), the SmartHG research project (FP7-ICT-2011-
8, ICT-2011.6.1), and the Spanish Ministry of Economy and Competitive- (e.g. line and switch statuses) provided to the estimator are
ness project RESmart (ENE2013-48690-C2-2-R). correct, ensuring that the network model is accurate and up

Graphical overview of the main functions provided by power system state estimation. buses. pseudo-measurements (i. [22]. The network state is expressed as the vector x. are not observable. or carrying out a Direct Current (DC)-only SE by neglecting all branch resistances and shunt z = h(x) − e (1) elements [5].g. measurements of active and reactive power flows in One of the problems encountered in general SE is the system branches. Each of the characteristics of distribution networks differ fundamentally ¯ are set according to the inverse of the variance weights in W from transmission networks in the following ways: . bad statistical testing. If the network.e. or ¯ −1 − H(ˆ Cr = W x) G−1 HT (ˆ x) (7) parts of it. such as the Weighted Least Average Value estimator and To estimate x. where ρjj is the diagonal of the covariance that sufficient measurement data is available for the SE. DSSE presents a number of new challenges. In order to reduce network quantities. since the where: W¯ is the measurement weight matrix. non-linear equations: fast decoupled SE [22]. e. H = δh(x)/δx.e. observability analysis is establishes rn = ρ−1jj r. some authors have proposed a forms a set of over-determined. which is normalised by to date [12]–[15]. the set of measurements from the network. The presence of bad data in the system measurement Figure 1. or any combination of the above. processing and removal are given in [17]–[19]. the methods and assumptions used where h(x) are the power flow functions corresponding to in transmission-level SE described above are often not valid each measurement in z. data set can be detected by applying statistical tests to the objective function J(ˆ x). where the J(ˆ x) Performance Index and data processing is used to identify and remove data affected Largest Normalised Residual Tests [7] are most commonly by gross errors and noise. This allows the weights in W ¯ to be adjusted so that the estimator gives more weight to input data points which are known to have greater accuracy. Next. Distribution System State Estimation ¯ (z − h(x)) Initial research into DSSE began in the 1990’s [24]– ⇒ minx (z − h(x))T W (3) [27]. Finally. the voltages and power angles at each node in the system. containing There are also alternative SEs discussed in the literature. These is applied. z. the Schweppe-Huber generalised M-estimator [3]–[8]. [21]. The minimisation in (1) is solved iteratively as follows: ∆zn = z − h(xn ) (4) ¯ −1 ¯ ∆z ∆xn = (HT WH) HT W n (5) ∆xn+1 = xn + ∆xn (6) where the Jacobian matrix. and e is the vector of measurement when considering distribution systems. i. estimates) of computational complexity of solving (2). The most commonly-used approach to minimise established techniques used in “conventional” SE cannot the objective function J(x) is the “conventional” Weighted be applied directly [23]. and to the normalised residual vector given by r = z − h(ˆ x). of the corresponding metered system measurement. estimated values of network inputs (often referred to as pseudo-measurements) need to Bad data is detected and identified (provided there is be provided. state estimators designed specifically for use in distribution networks. However. This has motivated research into Least Squares (WLS) method: DSSE. and n is total number of SE iterations. due to measurement or used in conventional SE. Further studies into bad data communication system failures [17]–[21]. SE uses the available measurement data to sufficient redundancy in the measurement data set) through find a unique solution for the system state. This the computational burden. The matrix: observability can be quickly determined by examining the null space of the Jacobian matrix [16]. minx ¯ − h(x))2 J(x) = W(z (2) B. but of power/current injections or voltage magnitudes at system otherwise the overall approach to SE remains the same. and many of well- errors. The values in z can comprise of measurements replace J(x) in (2) with a different objective function.

nature. requiring the use of full three-phase system model the system using the state-space form introduced models. to the construction of distribution systems (radial feeders and used in the EKF equations. “dynamics” in power systems is strongly associated with sume that the network is a balanced system. DSSE relies on pseudo-measurements of the demands or parameter errors [37]. synchronised metering devices. While most of the FASE tech- at each load point in the network. due matrix Hk is evaluated with the current predicted states. [26]–[28]. In addition. based on historical data niques and applications proposed to date are focused at the or load forecasts.g. rather than over-determined. such as Phasor Measurement 3) DSSE in Unbalanced Networks: In [25]. How. (e. the number of if the new measurements are significantly different from telemetered devices that can provide system measurements is the predicted values [39]–[41]. This has the advantage that Since distribution systems are typically very large and the Jacobian matrix H can be decoupled on a per-phase dense. available. particularly in dealing with noisy input data and wk and vk representing the process and observation and “robustness”. (e. for DSSE need to be scalable. DSSE is the computational complexity. gk is a vector limitations to adapting approaches from transmission-level representing the trend behaviour of the state trajectory. Forecast-Aided State Estimation at the distribution level are measurements (or pseudo- measurements) of power or current injections. A short-term forecast (e. FASE approaches are also interesting for than actual measurements. have a relatively low This approach has been called “dynamic” SE [36]. i. • Measurement types: Most of the available input data A. in that the estimation of the system state is only • Scale and complexity: Distribution systems are diverse dependent on the current “snapshot” of input measurements. allowing conventional SE methods to be applied to nodes) one of the most challenging aspects of implementing distribution systems which are unbalanced. techniques and applications of DSSE in the literature. in [37] and the Extended Kalman Filter (EKF) [38]: Some of the techniques developed in order to overcome these issues are discussed below. which have significantly lower accuracy transmission level. • Redundancy: For technical and economic reasons the number of measurement points in distribution networks III. the Jacobian in the presence of gross input errors [29]. Systems This section briefly describes some of the most advanced are under-determined. three-phase DSSE techniques [34]. and each time a new set of measurements becomes not work when applied to DSSE [30]. S TATE OF THE A RT is much lower than in transmission networks. the ability of the estimator to reach a noises which are assumed to correspond to white Gaussian unique solution of the minimisation described in (2) and (3) noise with zero mean. in which network branch currents. and high R/X ratios). current-based SE methodology was developed. In many anomalies such as bad input data. or bad data identification. can have significant phase concepts in FASE is given in [9]. An excellent summary of the main ever. At each time step k. Most FASE approaches imbalances. and network configuration cases. There are also SE those in urban areas) and have very large numbers of techniques which are designed to recursively update the state components. SE to DSSE. rather than node voltages are used B. however computational burden.e. often with high R/X ratios. comprising of many thousands of individual basis. networks in rural areas are very different from and not on previous input data values. allowing detection of of the entire network. ferred by many authors to avoid confusion since the word • Phase imbalances: Conventional SE techniques as. Multi-area and Hierarchical DSSE Techniques to represent the system state x. or have single.g. Load estimation techniques and DSSE. Units (PMUs) [42]. an “innovation analysis” can be used to determine 2) Load Estimation for DSSE: In DSSE. the fast decoupled methods and DC several seconds/minutes ahead) of the state variables is approximations often applied in conventional SE simply do made. This analysis filters the new often very limited. transient stability studies. In conventional . [43]. However. [35]. there are significant where Fk is the state transition matrix. A number of papers have construction (whereas transmission systems are more built on this approach in order to develop robust and accurate meshed). particularly if high-resolution data is available from their application to DSSE are discussed in [31]–[33]. and not sufficient to allow observability input data using the EKF equations. a branch.g. distribution systems. xk+1 = Fk xk + gk + wk (8) 1) Adapting Conventional WLS Techniques to DSSE: Many of the earlier research papers on DSSE focussed zk = hk (xk ) + vk (9) on adapting conventional WLS techniques to distribution networks [24]. This means that the methods developed estimate in order to track changes during normal operation. Direct The SE methods discussed in Section II are static in measurements of voltages and power flows are rare. • Construction: Most distribution networks have a radial phase or two-phase lateral feeders. and be applicable across a range the term “Forecast-Aided State Estimation” (FASE) is pre- of different network types.

or In [63]. However. [51]. insight into the behaviour of end-user loads. M . e. a machine learning approach is used to develop hierarchical SEs. The advantage of this approach DSSE [48]. a multi-area method has been developed for a range of power systems and smart grids applications.e. the relevant DSSE is func- DGs. authors have investigated the development of multi-level. such as model identification [67]. A information flow. leading to improvement of the performance of the DSSE Due to the need for better situational awareness and over time. work in this area. While the are significant opportunities for further tion carried out locally (i. In [62].g. made into the incorporation of smart meter data into DSSE It is likely that future DSSEs will make extensive use in [56]. F UTURE R ESEARCH A REAS FOR DSSE automatically detect new connections (e. In [46]. to distribution systems [50].. is the local state vector of mea. However. there has been much interest of “closed-loop” DSSE methods. of event-triggered approaches. Without the automation of loads is becoming available. Further research is required on the implementation more active system support. DG units. estimators capable of integrat- ing both analogue and digital inputs (e. all measurements are typically processed in low data reliability. and the lack of smart meter data time- one centralised SE. and several point power injections for use in the DSSE algorithm. allowing to improve load estimation techniques. a suspicion of a network topology for a number of reasons. In the multi-area approach. e. the scope for using smart meter data only when the measurements received indicate that there is as an input to DSSE for real-time applications is limited a potential issue. [65]. 15-minute or hourly intervals). and data Computational Intelligence Methods in DSSE is exchanged between areas only where they border each Computational intelligence methods have been proposed other. particularly with regard to voltages. [49]. the DSSE is carried out computational effort. the SE is solved locally within each measurement area. is presented in [66].g. a closed-loop DSSE allows the predictive number of studies have investigated “Advanced Distribution database used to estimate loads (and DG outputs) to be Management Systems”. power/voltage where xm = [xim xbm ]. This approach is particularly important in the context smart meter data rates (e. . A key feature of future DSSEs will be the flexibility to incorporate multiple zm = hm (xm ). for the total number of SCADA. in the network area of interest). DSSE. where DSSE is an the SE [64]. DSSE accuracy. or “measurement areas”. This data can be used to better the majority of control and network model management understand and model the behaviour of distribution network functions. low error. smart metering. a neural network for communication and interaction between transmission approach is used to create pseudo-measurements of load and distribution network management systems. border state variables xbm . An “autonomous” approach to DSSE important part of the methodology [51]–[55]. extensively in the power systems research literature for load oped separately. for DSSE. which contains the internal state variables measurement data from a range of diverse sources.g. previously only used at SE methods proposed in the literature have an open-loop the transmission level. PMU. (10) types of input data... In [59]. However. systems designed to optimise energy continuously updated and improved based on feedback from management in distribution networks.g. The multi-area a certain degree of visibility in parts of the distribution SE can be expressed as [47]: system which were previously unobservable. including: data privacy issues. Advanced Distribution Management Systems themselves as more measurement data becomes available.g. Traditionally. measurements and switch/breaker statuses [60]) and also surement area m. which meets the performance requirements for Machine learning techniques are particularly attractive for real-time applications in very large networks. from DG units) to the distribution network and update the system model Application of Advanced Metering Infrastructure Data accordingly. [57]. and distribution networks in a DSO region would lead to an ultimately. there are potentially significant distribution systems may be to split the networks into a benefits in terms of applying smart meter data for providing number of smaller sub-networks. measurement areas. designed to integrate transmission SE and load estimates for DSSE. and losses in the low voltage distribution network carrying out advanced DSSE functions requiring significant is demonstrated in [58].g. m = 1. [68].SE methods. where the aim is to identify . of DSSE model identification. Some initial studies have been unreasonable increase in operator workload. is that load models are designed to recursively re-train C. While almost all of the in adapting operational techniques. The further application of machine learning The widespread introduction of smart meters means that techniques will be crucial in allowing DSSE to be im- an unprecedented amount of detailed historical data on user plemented on a large scale. The use of smart meter data to estimate flows. a better solution for large synchronisation. Artificial neural networks have already been used transmission-level and distribution-level SE have been devel. and in providing in which the SE is solved locally [44]–[46]. the implementation of DSSE across all of the loads. xim . e. using “compressed” measurements from smart meters and In the event-triggered approach. in which the DSSE is designed to IV. M. with the increased requirements estimation and forecasting [61].

IEEE Trans. pp. Liu. Liu and S. 1069–1075.. 24. 1. 2. 3. “Fast decoupled state estima- vehicles. and carrying out a Monte-Carlo simulation. 2. vol. 15. 1645–1652. 2009–2015. algorithm works by assigning an equal initial probability [13] I. IEEE Trans. 4. 7. 666–678. vol.” Proceedings the true network configuration using a Recursive Bayesian of the IEEE. Leao. I. N. pp. of future distribution networks. configurations that are critical to the DSSE are stored. 1990. updating (11) at each iteration.-H. May 2012. Abur and A. and A. 5. “Choice of estimator for distribution system state estimation.” Power Apparatus & Sys. 101. Monticelli. . discussed in this paper in detail. vol. Feb 1992. Wu.. “Power system parameter estima- tion: a survey.-H. vol. Gomez-Exposito. vol. Stott. Garcia. [24] I. as these systems become pp. pp. “A survey on state estimation algorithm no. Journal of Model identification is particularly relevant in distribution Elec. IEEE Trans. 2004. M.. and Y.” Power Sys.. the model which has the highest probability is selected as [7] A. given [12] W. Power & Energy level. Brown Do Coutto Filho and J. Jabr. using a Gaussian distribution as in [67]. A. Smart Power Grids. 33–43. W. and X. pp. k k−1 ) j=1 p(ej | ηj ) p(ηj | ε [10] A. Oct 1993. 1995. are typically unmonitored. 14. “Bad data communications technology infrastructure at the distribution detection and identification.” Int. Nov 1993.. Schweppe and D. demand-responsive loads. 98. 94 – 103. Norwell. 1667–1677. “State estimation applications related to state estimation in electric power observability based on the null space of the measurement jacobian matrix.” Power Apparatus & Sys.” Power Sys.-H. where the switch statuses throughout the network Approach. Huang.. state estimation.-J. estimation in electric power grids: Meeting new challenges presented p(ηi | εk−1 ) is the prior probability. in which all of the network and Implementation. IEEE Trans..” Power Sys.. vol. [15] O. no. ε where = [11] Y. electricity storage. Nov 2006. Vempati. MA: Kluwer. pp.” Power Apparatus & Sys. V. 5. 1. The Power Sys. Zarco and A. on the results of the DSSE applied to each network model. New York: Marcel Dekker. IEEE Trans. and S. Wu. no. B. approach. and K. 81–89... Feb 1995. Aug 1990. IEEE Trans. Feb 1995. S. k ch. Feb 2000. 5.” calculated. G. pp.. 1126–1139. pp. Kashyap. Teng. Journal of Elec. no. 950–961. Rom.” Power Sys. pp. 3. Misra. quickly and reliably. pp. no.” Power Sys. 12. C ONCLUSIONS Aug 1998.. in order to recommend the [23] R. 10. Beiler. Kelley. 1. vol. detection and identification of multiple bad data in power system state estimation. Power & Energy Sys. “Power system static-state estimation: Parts Delivery. 1. and microgrids. small. Singh. and R. May 1983. 1656–1658. e2 . “Estimation of parameter errors model ηi is correct. pp. vol. no. pp. Garcia. eN is the set of error vectors of each model. 4. Koglin. Pandey. “A branch-current-based state estimation 125–130. & III. 262–282. much more active and complex due to increasing pene. Huang. DSSE has not been widely [18] A. “Electric power system state estimation. 8. Power System State Estimation: Theory development of a model bank. However. 1608–1615. 2012. no..-R. pki can be from measurement residuals in state estimation [power systems]. pp. Spring Congress on Engineering and [26] C. vol. 1–4. 80 – 87. 439–478.. Further work tion and bad data processing. Aug distribution networks. Conejo. 2010. R. 229–240. [3] F. 2..” Power Sys. IEEE. Leite da Silva. no. and P. S. 10. Expsito. IET. no. Chen and A. The a posteriori probability of the model ηi being [8] P. Power & Energy Sys. “Bad data pre-filter for trations of variable. 10. Feb 1982. “A new approach for mance. 13. 216–222.. no. F. IEEE Trans. Shahidehpour. 2012 IEEE. applying AMI data in order to improve estimation perfor. “State estimation for electric R EFERENCES power distribution systems in quasi real-time conditions. Jan 1970. Singh. no. This paper provides a detailed survey of techniques and [16] E. T.” in Proc. [25] M.-L. . vol. gence methods to DSSE. Despite the high level of research pp. Springer. best approaches for future large-scale implementation. 12. Werner. Power Sys. G. no. method for distribution systems. no. it is becoming more and more evident Sys. Filho. vol. “Identification of topology errors in power system state estimation. will examine the adaptable and intelligent DSSE approaches vol. no. pp. “Power system state estimation: a survey. Abur. 1. Sept 1979. vol. and the application of novel computational intelli. and R. and C. Castillo. pp. V. 3.” Power Apparatus & Sys. Nian-de. F. interest in the last two decades. Pruneda. 4.” Signal Processing Magazine. IEEE Trans. of distribution grid. The a priori probabilities. Falcao.. highly-dispersed resources. It also discusses the potential for 2005. Lun. IEEE Trans. IEEE Trans. Shiyu. “Reliable bad data processing for real- implemented in distribution networks to date. 29. & Dist.. Pal. “Placement of PMUs to enable bad data detection in state estimation. pp. A. no. Arefi and M. to each model. Gupta. Monticelli. Tran. IEEE Trans. C. Lu. 10. 89. conditional probability of the error in the ith model. Monticelli.  k k N kis the total number of models. IEEE Trans. IEEE Trans. and V. pp. J. J. IEEE Trans. Alsac. 2. vol. pp. 3.g.” Power [1] F.” Power Sys. State Estimation in Power Systems: A Generalized systems. 20. “Bibliography on power system state estimation (1968-1989). 1531–1538..-F. N. and W. B.. It is demonstrated in [67]. correct is given by: Feb 2000.” Int. Costa and J.” Power Sys.. pp.the correct network model from a range of possibilities. and p(eki | ηi ) is the by the requirements of the future grid.” Int. scale renewable energy.. Zhi. Based [6] M. vol.” Gen. that this approach can converge on the correct model mation in power system state estimation. that DSSE will play an important role in the operation [20] J. II. no. vol. no. 454–462.. Min.” Power Sys. Solares. Journal of Elec. [14] W. “Parameter error identification and esti- [68]. This has time state estimation.. 4. no.. e.. “Distribution system state estimation. 1999. “Forecasting-aided p(ek | ηi ) p(ηi | εk−1 ) state estimation: Parts I & II. 32. electric [22] A. 8. [2] Z. 5. 483–491. R. L. vol. The authors in [67] propose the [5] A. vol. vol. Neisius... no. vol. IEEE Trans. Baran and A. no. De Souza. pp. Nov 2009. [4] A.. Shi-ying..” Technology (S-CET). Singh. and D. Monticelli..-M. p(ηi | εk ) = PN i (11) pp. 1990. [17] X. vol. 102. IEEE Trans. pp. no. 1165 – 1174. no. Monticelli and A. vol. [9] M. Abreu. “Generalized state estimation. “State e1 . Er-keng.. 21.. A. mainly been due to a lack of investment in information and [19] H. pp. Haghifam. 1.. Schmitt. 2009. 200–209. Roytelman and S. 2011.. [21] D. 88. Lim. Liu. 7.

9. IEEE Trans. “Design of smart distri- through gaussian mixture model of the load as pseudo-measurement. 12. IEEE Trans. Pal. [68] Y. Baran and T. Salazar. 2. Krogh. 375–382. [62] E. and M. IEEE Trans. Journal of Elec. June 2014. June 2012. IEEE Transactions P. 28. Souza. Italy. A. “A multilevel state estimation paradigm for smart grids. Kwon. [65] B. Shih and S. vol. Kekatos and G.. [58] G. vol. 1617–1626.” Elec. and C. E.” Power Sys. 43–48. 28. and G. Choi. Smart Grid. [64] B. J.” Electric Power Systems Research. pp.” estimation for power distribution systems. Nov 2012. 18th Int.” in Power Systems Conference and Exposition. “Unscented filtering and nonlinear estima. 1–6. 4. no. Jul 2001.” in Proc. IEEE Trans. E. de la Villa Ja´en. vol. IEEE Trans. vol. Power Sys. “Optimal power flow for maximizing network benefits from demand- “Control of active networks.” Power Systems.. Dzafic. vol. [53] D. pp.” Power Sys. IEEE estimation through kalman filtering and load forecasting. vanced distribution management system. R. Thornley. Gruber.” Power Delivery. 5. Power Sys.” Power Sys.” Smart Grid. Natarajan.” Smart Grid. [32] A. no. pp. 1. Pau. vol.. M. & Dist. pp.” Instrumentation and Measurement. Ghosh. Lee. Giannakis. Strbac. 4. vol. vol. and G. no. IEEE Trans. Cutsem. pp. Power tion by measurement pairing. F.. Zhang. Yun. “Short-term operational planning and no. Weaver. IEEE Trans. Dec 2013. 877. Kwak.[27] A. Huang. and B. DMS applications. and control. no. [49] A. 4. B.. no. and D. 50–59. 2014.. Moreno. no. Sun. 667–672.” Power Sys. pp. R. B.-M. pp. Lopez. 27. Gilles. approach and its applications to power systems. “Distribution system state estimation [54] I. IEEE Trans. pp. Teng. He. Mar tures. & Dist. [56] M. pp. 1. 1739– Distribution. vol. vol.. Meliopoulos. “A distributed multiarea state estimation. Meliopoulos. 9. Hayes. no. K. Meliopoulos. Rousseaux. pp. Jenkins.-J. 4. Tran.. C. [66] S. 3. Hayes and M. no. Mei. Ilic. pp. no. 16. estimation methods with the use of AMI metering for distribution [37] A. 81. no. “Investigation load no. 2014. and C.” in CIRED 2014 [44] G. Manitsas. “A seidel-type recursive bayesian 1060 – 1069. vol. Apr 1997. no. 6.” in Rural Electric Power Conference (REPC). 89. no.. Sept 2013. Maloyd. Strunz. “A robust three-phase [55] S. 2011.” Gen.” Power Sys. Yu. Manitsas. “Real study: Autonomous state estimation in distribution systems. pp. Julier and J. Pegoraro. Tran. IEEE [48] A. 1. “Ad- in treating current measurements for the branch-current-based distri. for security analysis applications.” Power Apparatus & Sys.” Power Sys. IEEE Trans. vol.. vol. B. 1187–1195. B. pp. “Real-time monitoring and economic [31] M. 1640–1655. pp. Jabr.” Proceedings of the IEEE. 433–439. 141–147. Singh. pp. no. and A. de la Villa Jaen. R.. 92. B. March 2009. 1–10. “Distribution fast decoupled state estima. estimation in integrated power systems. Sakis Meliopoulos and F. F. vol. Kelley. June 2005. “Analog-digital power [40] K. Power & vol.. Zhao. 330 – 336. Polymeneas.. J.. [36] A.. and A. 2.” Gen. 45. IEEE Trans. A. S. pp. and R. no. 4. vol.” Elec. pp. Jabr. 1631–1642. 3. Do Coutto Filho.. and N. and T. “A recursive bayesian 4839–4848.” dynamic state estimation of a power system. Haughton and G.” bution management system for obtaining real-time security analysis Gen. IEEE. P. Sept 1970.. A. July 2014. Abur.. Jones. Samper. Aug 2012. smart distribution systems. Debs and R... of a power system. [29] A. and R. 99. R. De Queiroz. and X. B. 1747. Nov 2013. pp. “State estimation for real-time monitoring dispatch of smart distribution grids: High performance algorithms for of distribution systems. J. Heydt.. in press. Teng. vol. March 1888–1896. G. D. [33] R. ments. 401–422. 26. 939–946. “Feasibility [46] I. Singh. Pahwa. “Distribution system state estimation Tran. pp.. Gao. and M. 2109–2117. no. and N. 0. and S. 2005 CIRED. “Load modeling for distribu. no. IEEE/PES. 1710–1711. no. branch-current-based state estimators for distribution systems. “Efficient branch-current-based voltage distribution networks. “Smart grid technologies for autonomous operation measurement modeling. 26. 28. Sept 2013. and T. “Distribution grid state esti- [39] D. vol. Surapong. 237–244.. June 2011. 28.” Smart Grid. pp. R. Shafiu. G. 2008 IEEE. M.Research. no. 1. 17.. vol. “A taxonomy of multi-area state on. and S. [52] A... R. Gomez. and predictive operation in Korea. Djoki´c. 2014. McDermott. Hippert. Jan 1996. 25. 6. Y. [60] H. S. and A. Lin. G´omez-Quiles. 1. pp. 2004. Dugan.. Nov 2011. Aug 2010. [34] D. IEEE Trans. IEE Proc. 1329–1336. A. “Power system operating state forecasting mation from compressed measurements. Pal. Hernando-Gil. Pal. January 2010. pp. no. 5. no. 7. S. IEEE Trans.-R. 3. Tan. & Energy Sys.. IEEE Trans.” Power Sys. Leite da Silva. 73–84. Choi. 999–1005. IEEE Trans. vol. and S. no. S.” Smart Grid.” Int. Prodanovi´c. Korres. 6. Cokkinides. R. pp. 16. B. Natarajan.. Sulis. –. 5. May 2013. “A robust state estimator for medium [43] M. A. G. 3. 152 – 158. B3–1–B3–9. H. F. Tran. no. 3. [30] W. Pal. Cokkinides. pp. May 1996.” Power Sys. vol. Zhang. and S. He.” Smart Grid. 2109–2117. Proceedings of the IEEE. no. Feb 2011. Research.-H. 2. Pahwa. no. “Distribution system [42] A.. Larson. Journal of Elec. Collin.. E. and R. Feb 2002. P. vol. no. Pedreira. Liu. 44–55. IEEE Trans. pp. V. IEEE Trans.. & Dist. R. Trans. bution state estimation. Singh. 2000. Aug 1994. Seifi. Uluski. 62. B. “Multiphase power flow and state Quiles. Workshop. July 2008. 1–4. M. 1670–1678.. 143... vol. Song. I. pp.. “Neural networks for short- [41] E. 2012 forecasting in electric power systems. distribution system state estimation including synchronized measure.” Smart Grid. IEE Proc. H. “A highly efficient algorithm [51] A. S. H. pp. 2013. 2. 130. and F. Farantatos. “State estimation and state estimation algorithm for distribution networks. Huang. Feb 2001. Blood. 2009. pp. IEEE Trans. no. 1–3. no.. toward smart grids through large-scale advanced metering infrastruc- tion.” Gen. 2419–2429. 2. Falco and U. 4. N. “A new coordinated approach to state [28] W. no. Lubkeman. 1008–1016. pp. IEEE Trans. vol. “A closed-loop state esti- [45] V. Vargas and M.” Power Sys. Deshmukh. state estimation using an artificial neural network approach for pseudo Y.. state estimation in power distribution networks. vol. 3. V. Henselmeyer. F. IEEE Trans. Wu. “Application of a robust algorithm for system state estimation based on information theory: Parts I & II. Z. tribution system state estimation. G.” Int. G´omez-Exp´osito.. “A dynamic estimator for tracking the state PSCE ’09.” Power Sys. on Electricity side management. Amaras. “Paving the road [38] S. Jan [35] J. Hayes... “A linear state estimation formulation for tion circuit state estimation. Uhlmann.. Kim. May 2013. and S.” Power Delivery.” Power time estimation of loads in radial and unsymmetrical three-phase Sys. March 2013. Voltage/VAr control in distribution network with intermittent mea- surements. Lin and J. vol. 1. approach for identification of network configuration changes in dis- [47] A. 149. and J. vol. Strbac. 200–209. pp. Abbasi and A. [57] R. 1991. Thukaram. vol. 29. pp. no. Teng. 2. Bezerra. 3. September 1983. Baran and A. Jenkins. no. 4. vol. [67] R.. pp. pp. vol. 27. 4. “State system analysis. 2011.” Power Sys. C. no. 866– 1601–1609. J. G.-J. pp. “Distributed robust power system state mation tool for MV network monitoring and operation. Chen. no. pp. Chen. April 2012. Harrison. Jerome.. 1.. pp. 191 – 200. no. IEEE Trans. 3.-Y. IEEE Trans. 4. IEEE Trans. and Energy Society General Meeting. and A. 55. pp. IET. “Using voltage measurements to improve the results of 2014. vol. IEEE Trans. pp. 1.. 1.. vol. 952–976. Strbac.” in Power Trans. Gomez-Exposito. Conf. . Arritt. distribution networks. P. Nov 2002. pp. Prodanovi´c. vol. [59] S. “Electric power system static state term load forecasting: a review and evaluation.. 4. vol.. vol. no. [50] B. Huang. & Dist. 11. [61] H. estimation.. Energy Sys. 4. using AMI data. IEE Proc.” Power Sys. Rome. [63] J. pp. vol..” Smart Grid. 13. Alam. 2. estimation methods. May 2013.