Research On Optimal Reactive Power Dispatch In Active Distribution Networks With High Penetration Of Renewable Energy Sources

The high penetration of renewable energy sources has become a prominent feature in future distribution network.As featured of stochastic and intermittency,the renewable energy sources bring uncertainty,bilateral power flow and other factors to distribution network,which may lead to power supply reliability reduction,power quality deterioration,economic and safety of operation reduction and so on.The active distribution networks(ADNs)control the controllable resources actively to support system operation,resist the adverse effects brought by renewable energy sources and guarantee the system safe,efficient and stable operation.The optimal reactive power dispatch(ORPD)plays a core role in implementing the active control and management of distribution network.The researches on ORPD in ADNs are conducted in this paper while the main work is as follows:In the first place,an adaptive robust optimal reactive dispatch method is proposed.In order to alleviate the shortage of traditional robust optimization(RO)as conservation,the idea of adaptive robust optimization is inverted and the adaptive function is constructed based on affine mathematics.The approximate linear relationship between node power injection and node voltage amplitude is obtained by the approximate linearization of power flow model.Based on this linear relationship,the adaptive robust voltage constraint is built by transforming the voltage constraint into power injection constraint.The nonlinear adaptive robust distributed generation(DG)constraint is constructed and transformed into a set of linear inequalities by a polygonal inner approximation.The adaptive robust reactive power dispatch(AR-ORPD)mathematical model of three-phase balanced distribution network is constructed based on the distribution load flow(Dist-Flow)while the AR-ORPD mathematical model of three-phase unbalanced distribution network is established based on the node injection power model and the semi-definite programming.Finally,the cutting plane solution method is given.The method of utilizing adaptive function to replace the conservative solution also realizes the adaptive adjustment of DG's reactive power injection based on its active power real-time disturbance,which provides an idea for the real-time control of DG in the subsequent research.Secondly,the multi-timescale robust voltage/var control(MTSR-VVC)strategy which consists of “hours-level”,“minutes-level” and “seconds-level” is proposed.In this strategy,the traditional devices as on-load tap changer(OLTC),capacitor bank(CB)and the novel devices as DG is controlled coordinately.OLTC and CB are dispatched in “hours-level” while DG is scheduled in “minutes-level” and “seconds-level”.The box uncertain set is utilized to describe the uncertainty of DG's active power,active and reactive load demand.The two-stage robust ORPD model which controls OLTC,CB and DG inverter coordinately is constructed based on the second-order cone programming.For the real-time control in “seconds-level”,the Q(V)model aiming at mitigating voltage fluctuations and the Q(P)model aiming at economic optimization are given respectively.As the reference solution of “seconds-level” is obtained from the “minutes-level”,the coordination of the three different levels is realized.The columnand-constraint generation(C&CG)algorithm is utilized to solve the two-stage ORPD model.The power flow calculation based iteration algorithm and cutting plane algorithm are utilized to solve the Q(V)model and Q(P)model respectively.Thirdly,the framework of “distribution network partition-distributed optimization” is built.There are different coupling characteristics between different nodes in ADNs.If the nodes with high coupling are aggregated into the same sub-network while the nodes with low coupling are separated into different sub-networks,then the network is partitioned in a “high-cohesion and low coupling” pattern.Based on this type of network partition,different sub-networks are featured of high sparsity,then only rare information communication between sub-networks are required.Therefore,the computation efficiency of distributed optimization would be improved.The electrical distance is introduced to characterize the coupling property between different nodes.The network partition model is established based on the constructed partition indexes which consist of the “intra-sub-network cohesiveness index”,the “inter-sub-networks sparsity index”,the “sub-network count index” and the “physical connection index”.The virtual load is established to describe the load characteristic of its downstream sub-networks.Introducing this virtual load into the boundary conditions would improve the efficiency of information communication,then to accelerate the distributed computation.Based on the optimal network partition and virtual load,the distributed optimal reactive power dispatch(D-ORPD)model with the aim of reducing power losses is established and solved by the alternating direction method of multipliers(ADMM)algorithm.Fourthly,the distributed multi-timescale robust optimal reactive power dispatch(DMTRORPD)model and its solution method are proposed.ADNs are featured of large scale,complex operation status and flexible network topology,then the centralized operating and controlling method faces enormous pressures in computing resources,storage resources and global communication.Considering this factor and the uncertainty,the DMTR-ORPD model is constructed based on “distribution network partition-distributed optimization” framework and the second-order cone programming.A novel network optimal partition method based on the affine propagation clustering algorithm is proposed.In this method,the network is partitioned optimally based through its own learning.Thus,this method is well applicable in engineering.The uncertainty is expressed by the polyhedral which is conservation adjustable while the “virtual load” is introduced to improve the information communication efficiency.The distributed model is solved by the integration of ADMM and C&CG algorithm.The distributed computation framework is given by ADMM while the regional two stage robust optimization problem is solved by C&CG.Considering the difficulties for guaranteeing the convergence of ADMM in solving models with discrete variables,the alternating optimization procedure(AOP)is introduced to improve the distributed algorithm.Then,the DMTR-ORPD model is solved effectively.Finally,case studies are conducted on the IEEE13,IEEE33,IEEE123 and PG&E 69 test systems to demonstrate the validity and effectiveness of the proposed methods in this paper.