Stochastic Optimizaiton Method For Power Management In Active Distribution Netwroks With Photovoltaic Generation

Power system optimal power managements can ensure the economic operation of the power system.As the power system advancement,active distribution network(ADN)with high penetration renewable energy resources will be the key part of the future power grid.The photovoltaic generation(PV)unit is one of the renewable energy resources.However the stochastic nature of solar renewable power leads to unexpected variations in electricity generation levels.Abrupt solar irradiance condition changes may lead to inadequate or at times excess generation,which eventually may result in voltage fluctuations across the distribution feeders.These present a challenge in optimal power management of the active distribution network.To cope with the random and intermittent nature of solar generation,a stochastic optimization model for real and reactive power management in such ADN with a large number of residential-scale PV units is introduced in this thesis.Firstly,the fundamental structure of ADN as well as the controllabilities of the consumption power of the loads,reactive power consumption or generation by the PV inverters,etc.are analysed.Power flow equations of the ADN are formulated.A linear approximation of power flow equations is obtained.The inequality constraints equations used for the optimization power management with a second-order cone programming relaxation method are modeled.The power injection models of the nodes with controllable loads and/or PV units are formulated.Secondly,the stochastic active power output of PV units is modeld using scenario analysis method.The optimal number of scenarios is generated using Wasserstein distance metrics.The objective function considered of the customer satisfaction,the expected cost of providing real power to opetate the ADN and the aggregate losses across the ADN is constructed.The overall optimization problem model is formulated using the power flow equations of second-order cone programming relaxation method.A centralized optimization solution methodolody requires communication of all local quantities to control center,which can be prohibitive once the network and/or the number of scenarios is considerably large.A decentralized algorithm based on the alternating-direction method of multipliers has been developed to solve the resulting optimization problems on a ADN,whereby only neighboring users will need to exchange information which entails limited communication overhead.Thirdly,a risk-averse formulation is presented for joint voltage regulation and power management.A conditional value-at-risk objective per node is incorporated to account for the risk of having absolute voltage deviation,in addition to terms corresponding to customer dissatisfaction,power generation costs,and line power losses.Finally,numerical experiments are conducted to evaluate the effectiveness of the formulated stochastic optimization algorithms and the decentralized solvers.A single feeder 50 nodes ADN and a 15 nodes ADN are used for the numerical tests.