Decentralized multi-agent coordinated secondary voltage control of power systems

Abstract

In this thesis, two different approaches toward Secondary Voltage Control of large scale power systems are presented. In the first approach, for each area of the power grid, a Model Predictive Controller which modifies the set-points of reactive power compensators participating in Coordinated Secondary Voltage Control algorithm is designed. The proposed controller takes into account reactive power limits of these compensation devices. The novelty of the method lies in the consideration of measured reactive power deviation on tie-lines between neighboring areas as measured disturbance and compensation of the disturbance by regional MPC controllers. As another contribution of this work, the validation of the proposed algorithm is done in real-time simulation environment in which the decentralized MPC controllers are run in parallel on separate computational cores. The stability and robustness of the presented algorithm is validated for a large scale realistic transmission network with 5000 buses considering standard communication protocols to send and receive the data. Simulation results show that the proposed method can regulate the voltages on the pilot buses at the desired values in presence of load variations and communication delays. The computational burden of the proposed method is also evaluated in real-time. For the networks facing large disturbances, an alternative model based centralized controller is presented next which considers the nonlinearities of the power system while taking into account both discrete and continuous type compensators. In this regard, sensitivity analysis is used to first find the most sensitive buses of the network called pilot nodes and second to locate the control buses in which discrete type or continuous type controllers are installed. The CSVC controller is then designed based on the notion of nonlinear sensitivity model which relates reactive power injection/absorption or change of reference voltage of controllers to the voltage variation at pilot buses at different operating points of the network. The non-linear sensitivity model is identified using Neural Networks approach which is then used by Simulated Annealing optimization algorithm to solve a mixed discrete-continuous type optimization problem and find the suboptimal control input. The proposed algorithm is tested in real-time against coordinated secondary voltage control method based on linear sensitivity models and also traditional capacitor/inductor banks’ control method which is based on local measurements. Finally, the same methodology as nonlinear sensitivity based optimal controller is adapted to a decentralized architecture considering consensus between regional controllers overlapping in some buses with a connected reactive power compensator. The consensus is reached in two iterations and does not require any communication link between regional controllers. Moreover the proposed method gives the flexibility to the shared compensators as agents to decide on their degree of participation in SVC algorithm of each neighbor based on their own performance objectives.

Date	31 May 2018
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Ouassima Akhrif (Supervisor) & Maarouf Saad (Co-supervisor)