              ! " OBJECT-ORIENTED MODELING OF A POWER NETWORK FOR MODEL-BASED VOLTAGE CONTROL Francesco PETRONE Marcello FARINA Antonio GUAGLIARDI Politecnico di Milano – Italy Politecnico di Milano – Italy RSE, Italy petrone.fr@gmail.com farina@elet.polimi.it guagliardi@rse-web.it Carlo SANDRONI Riccardo SCATTOLINI Anna VENERONI RSE - Italy Politecnico di Milano – Italy Politecnico di Milano, Italy carlo.sandroni@rse-web.it scattoli@elet.polimi.it annaveneroni87@hotmail.com ABSTRACT This paper deals with the design of dynamic regulators for voltage control in Medium Voltage feeders with Distributed Generators, which can be used as active control elements. A modular object-oriented simulation environment has been developed in Matlab® /Simulink/Simscape. This tool has been used for the design and validation of a control structure made by control loops acting at different levels. Specifically, two control structures are synthetized: the first one is made by standard PI-PID regulators, while the second one relies on the Model Predictive Control (MPC) approach. INTRODUCTION The liberalization of the energy market and the widespread diffusion of distributed generation rise new problems in the management and control of medium voltage (MV) distribution networks. In fact, the high variability of renewable energy generation can result in significant modifications of the voltage profiles along the open-ended radial feeders typical of distribution networks. In turn, this can produce unexpected bi-directional variations in network power flows, with severe consequences on the quality of supply. For these reasons, coordinated voltage control in distribution systems is becoming of paramount importance and has stimulated many research efforts, see e.g. [1]-[3], just to mention a few. Voltage control usually relies on the use of on-load tap changer and switched shunt capacitors, operated in a decentralized setting, i.e. with local control laws without high level global coordination. More innovative approaches are based on the direct use of distributed generation by directly exploiting the possibility of synchronous distributed generators to control their terminal voltage by adjusting their reactive power, see again [3] and the references quoted there. The corresponding control structures can be either fully decentralized, i.e. any distributed generator is controlled by a local control law, or designed according to a centralized structure. In any case, i.e. either for centralized or distributed control structures, the common assumption of sinusoidal regime cannot be advocated, and a rational controller design requires the knowledge of a (linearized) dynamic model of the system. In view of these considerations, the research activity described in this paper is aimed at developing new model- based distributed control methods for voltage control in MV feeders. The overall control system is composed by the joint use of a centralized regulator, which defines the proper settings of the reactive power along the feeders and of local decentralized regulators, one for every controlled distributed generator. In order to compute the required linearized models of the system and to test the control scheme in dynamic conditions, the first step of the research activity has concerned the development, in the Matlab®/Simulink/ Simscape environment, of a modular dynamic modeling and simulation environment allowing for the definition and simulation of feeders with arbitrary configuration. This simulation environment and the computed linearized model are first used to design a control system made by standard PI-type regulators where the dynamics is explicitly accounted for. Then, a Model Predictive Control (MPC) regulator is used to provide more flexibility to the control scheme and to improve the overall control performances. THE OBJECT-ORIENTED SIMULATION TOOL The typical elements of the feeder, i.e. synchronous generators, passive loads, transformers, transmission lines, asynchronous motors, have been modeled by means of the Park’s transformation, see [4]. The corresponding simulation blocks have been developed in the object- oriented simulation environment Simscape, a toolbox of Matlab®/Simulink for a-causal modeling and simulation. This choice stems for the need to consider systems described by DAE (Differential Algebraic Equations), and to connect them according to an a-causal configuration, typical of electrical networks. In the proposed simulation environment one can fully take advantage of the many features of Matlab/Simulink, such as the possibility to automatically derive the linearized models to be used in the control synthesis phase. Moreover, it is possible to link Simscape and Simulink models, so as to validate in