264 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 1, FEBRUARY 2005 Substation Data Validation by a Local Three-Phase Generalized State Estimator Antonio de la Villa Jaén, Pedro Cruz Romero, and Antonio Gómez Expósito, Fellow, IEEE Abstract—This paper presents a framework allowing the huge amount of data currently available at substations to be locally pro- cessed and filtered before being transmitted to the Energy Man- agement System. The proposed approach is based on the applica- tion of a three-phase state estimator (SE) at the substation level. In order to check the substation topology, individual switching el- ements are modeled following the ideas recently introduced by the generalized SE paradigm. Such a detailed model allows in turn the use of many more three-phase sets of measurements, like those pro- vided by the protection system, leading to a significant increase in local redundancy. Achieving this goal is facilitated by exploiting the possibilities offered by the so-called intelligent electronic devices, integrated in modern networked architectures which are becoming common in more and more substations. This way, both analog mea- surements and positions of switching devices can be locally vali- dated. This paper focuses particularly on new types of measure- ments which are included in the local SE formulation and presents experimental results showing the potential benefits that can be ob- tained when the proposed scheme is adopted. Index Terms—Generalized state estimators, intelligent elec- tronic devices, measurement, substation models. I. INTRODUCTION S TATE ESTIMATORS (SEs) determine the most likely state of a power system from sets of measurements which are captured remotely at substations and are collected periodically by SCADA systems. The estimated state is employed in en- ergy management systems (EMS) by a diversity of applications dealing with the operation and security of transmission net- works. The nonlinear equations arising in the process are solved by iterative algorithms, the convergence of which can be seriously affected by the presence of inconsistencies in the information coming from substations, like topology errors. In order to prevent such a problem as much as possible, SEs are preceded by a prefiltering stage that rules out certain raw measurements which are clearly wrong. This way, many topology and measurement errors can be detected and elimi- nated before actually performing the state estimation. The data validation proposed in [1] resorts to extra power flow variables that model the status of switching elements at the sub- station level. Active power flow measurements within the sub- station are used and approximate lossless models are adopted for power transformers. This yields a linear programming problem Manuscript received February 13, 2004. This work was supported by the Spanish MCYT and Junta de Andalucía under Grant ENE2004-06951/CON and Grant ACC-1021-TIC-2002. Paper no. TPWRS-00685-2003. The authors are with the Department of Electrical Engineering, University of Seville, Seville, Spain (e-mail: avilla@esi.us.es; plcruz@us.es; age@us.es). Digital Object Identifier 10.1109/TPWRS.2004.841176 whose solution allows topology errors to be identified in certain cases. Most proposals, however, centralize the prefiltering process at the control center, once the information from all substations is available. In [2], an approximate model is developed both for the active and reactive decoupled subproblems. Then, topology errors are identified by resorting to detailed substation repre- sentations. Only those measurements reliable enough are con- sidered in [3], and consistency checks are subsequently carried out on the remaining measurements. Artificial Neural Networks [4] or rule-based systems [5] have been also proposed for this purpose. The prefiltering scheme proposed in this paper is intended to be locally implemented at every substation where enough com- puting power and measurement redundancy allows it. Because of the relatively small problem size, it is possible to adopt a three-phase model in which all circuit breakers (CBs) are repre- sented in detail. This leads to a significant increase in the number of measurements available to perform the estimation, including those coming from the substation protection system. The main idea is to locally apply a three-phase generalized SE to obtain the substation state. Then bad data and topology error detection and identification techniques, developed in the last few years for generalized SEs, are applied to validate both analog measurements and CB statuses. The paper is structured as follows. in Section II a brief de- scription of modern substation automation systems is given. Section III describes the features of the proposed scheme. Next, Section IV presents in detail the resulting substation model, in- cluding measurements that have never been used before. Finally, Section V shows and discusses some test results, followed by the concluding remarks. II. SUBSTATION AUTOMATION SYSTEM According to [6], a modern substation automation system per- forms several local tasks with the aim of providing: • local and remote access to the power system; • local manual and automatic functions; • communications links and interfaces to the switchgear, within the substation automation system and to the net- work management system. These functions may be performed by smart multi-functional and communicative units, so-called intelligent electronic de- vices (IEDs), implemented nowadays by many electric utilities [7]. They are broadly defined in [8] as “devices incorporating one or more processors with the capability to receive or send 0885-8950/$20.00 © 2005 IEEE