Stochastic Volt/VAr control of power distribution systems
File(s)
Author(s)
Ul Nazir, Firdous
Type
Thesis
Abstract
The increasing stochasticity in the distribution networks is rendering the traditional volt/VAr
control (VVC) techniques more and more ineffective because of their basic premise that the
nodal net power injections are constant throughout the optimization horizon. This results in
voltage magnitude excursions outside the prescribed limits when the uncertain nodal injections
exhibit an appreciable variance. This thesis acknowledges the random nature of these nodal
injections, load and DG outputs, and seeks to achieve VVC solutions which are feasible for
most of the scenarios in real time. Thus, two novel VVC routines are proposed viz a centralized
algorithm based on chance constrained programming and a distributed algorithm based on
stochastic programming.
The centralized chance constrained VVC problem is solved through the scenario-based approach
which requires to collect large number of scenarios, making the problem computationally intractable, from the probability spaces of the random variables and ensure that the basic constraints of the original problem are satisfied for all of these scenarios. Therefore, an algorithm is proposed which only considers very small, but properly chosen, subset of the original scenarios for optimization, and at the same time guarantees that the basic constraints for every scenario are satisfied in the original scenario set.
The distributed stochastic VVC problem is solved by the alternating direction method of multipliers. This method allows the problem to be solved in a decentralized fashion by dividing
the whole network into various zones and equipping each zone with a local controller. A proper
coordination of the local controllers ensures the final solutions are globally optimal.
The developed algorithms are tested on the 95-bus UKGDS, IEEE 123-bus system and a 190-
bus system. The results show the superior performance of these methods by highly reducing
the voltage violations as compared to the traditional deterministic VVC routines.
control (VVC) techniques more and more ineffective because of their basic premise that the
nodal net power injections are constant throughout the optimization horizon. This results in
voltage magnitude excursions outside the prescribed limits when the uncertain nodal injections
exhibit an appreciable variance. This thesis acknowledges the random nature of these nodal
injections, load and DG outputs, and seeks to achieve VVC solutions which are feasible for
most of the scenarios in real time. Thus, two novel VVC routines are proposed viz a centralized
algorithm based on chance constrained programming and a distributed algorithm based on
stochastic programming.
The centralized chance constrained VVC problem is solved through the scenario-based approach
which requires to collect large number of scenarios, making the problem computationally intractable, from the probability spaces of the random variables and ensure that the basic constraints of the original problem are satisfied for all of these scenarios. Therefore, an algorithm is proposed which only considers very small, but properly chosen, subset of the original scenarios for optimization, and at the same time guarantees that the basic constraints for every scenario are satisfied in the original scenario set.
The distributed stochastic VVC problem is solved by the alternating direction method of multipliers. This method allows the problem to be solved in a decentralized fashion by dividing
the whole network into various zones and equipping each zone with a local controller. A proper
coordination of the local controllers ensures the final solutions are globally optimal.
The developed algorithms are tested on the 95-bus UKGDS, IEEE 123-bus system and a 190-
bus system. The results show the superior performance of these methods by highly reducing
the voltage violations as compared to the traditional deterministic VVC routines.
Version
Open Access
Date Issued
2019-10
Date Awarded
2020-02
Copyright Statement
Creative Commons Attribution NonCommercial Licence
License URL
Advisor
Pal, Bikash Chandra
Publisher Department
Electrical and Electronic Engineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)