#### Observability, fault detection and state estimation for large-scale interconnected power systems

**Encadrants :**Bodgan Marinescu

**Sujet :**

**Key words**: state estimation, transmission grid, robust estimation, grid protections, grid state-estimation

**Context:**

Estimation and detection are important tasks for power systems. Indeed, these topics appear in several contexts:

- estimation of the state of the grid, i.e., computation of a coherent and realistic operation point of the whole power system using uncomplete or not fully reliable measures. This can be on-line – for system operation – or off-line – for a posteriori studies of important incidents of the grid

- detection of a distant fault or of a disfunction of a power system in order to protect a local equipment. In this case a trip decision must be taken rapidly and by using thus only local information.

- detection of a change of the operation conditions of the power grid in order to switch to an appropriate control mode

The state of an electric power system is usually estimated by computing voltage magnitudes and angles for all grid busses. For this, measures are used from the grid along with some information on the generation. This information generally consists of active power P and voltage V outputs of the machines along with the power generation limits Pmax/min, Qmax/min. No information is used about the internal dynamics of the machines and the output of the grid state estimator consists thus of trajectories for the grid voltages sampled of quite large intervals (about 5min). As a consequence, this state estimation of the grid may not be compliant with the machines dynamics, especially under stressed conditions when some internal variables of the machines (like, e.g., rotor/stator current or state variables of the voltage or speed regulators) can reach their physical limits.

As a consequence, difficulties may be encountered to perform on line Dynamic Security Assessment (DSA) for which one need to initialize all the state variables of the dynamic model of the system, including internal dynamics of the machines. Usually, a set of Differential Algebraic Equations (DAE) is used for this and includes both equations of the grid and of the machines. The results of the state estimator of grid’s variables for a given time are used to compute the internal dynamic state variables of the machines assuming that the system is in steady state and in nominal operational mode. However, under the stressed conditions mentioned above, this initialization may fail or give wrong results unfortunately exactly when a DSA is required to help the operators in the control room. This problem can become even more frequent with power electronic devices equipped with discrete controls.

Most of the existing grid-state estimation approaches are adapted to deal with the very large-scale of the interconnected power-systems. However, recent reflection motivated by the evolution of the power systems due mainly to the increased penetration of the renewable energies and new power electronic technologies brought the power systems researchers to the concept of smart sub-station. These new type of sub-station uses a maximum number of local information in order to allow decentralised data treatment and rapid decision. It is interesting to exploit in this new context the benefits of recent robust and adaptive observation techniques already used in other applications context.

**Research subject, general work plan:**

- study the feasibility of an extended state estimation for which the internal dynamic model of the machines (electrical/mechanical parts and the regulators) are considered along with the grid equations in a coupled DAE model.

- large-scale implementation on a real power system.

- adaptation of estimation and detection techniques to the concept of smart sub-station

**Framework:**

This work is proposed in a general framework of collaboration with RTE – the French Transmission System Operator – and it is thus connected to real needs of the interconnected power systems. Realistic tests and validations of the theoretic developments mentioned above are possible on grid models and scenarios provided by RTE. The Control of Power Grids chair ( http://chairerte.ec-nantes.fr/ ) which exists between Ecole Centrale Nantes and RTE R&D guarantees the direction and the financial founding of this work. Several PhD subjects may be established for this general theme. The PhD work will be supervised in collaboration with specialists from Nantes and other research centres in France and abroad, according to the specificities of each subject.

The work will be carried out in Nantes-France.

**Equipes) encadrante(s) :**

- Commande

**Ecole Doctorale :**

**Prérequis :**