Research On Optimal Partition Control Strategy Of Voltage For Distribution Network With High PV Penetration

Under the influence of the power market and sustainable development strategies,distributed Photovoltaic(PV)technology has attracted worldwide attention due to its characteristics of flexible installation,environmental protection,and convenient power supply.The access of distributed PV makes the power grid change from the original radial network to a weak ring network with power nodes,which changes the original power flow.It will affect grid planning,protection,power quality,dispatch management,etc.After the original distribution network is connected to distributed PVs,the power flow has a great change.In particular,high PV penetration will cause reverse flow and violated voltage in distribution network.In order to solve the overvoltage,this paper proposes a double-layer partition voltage regulation strategy by optimizing the reactive and active power output of distributed PVs.The model uses the improved Fast-Newman algorithm to partition the distribution network,selects the priority regulation node set which contains nodes with distributed PV,and uses the CSO algorithm optimized by Transfer Learning(TL)to solve the optimal active and reactive power dispatching value of each node in the node set.The specific research is as follows :(1)In order to select suitable nodes with distributed PV for Reactive Power Compensation(RPC)and Active Power Curtailment(APC),an improved modular increment index is introduced in the Fast-Newman algorithm,which considered the PV access position matrix and reactive/active voltage sensitivity matrix.On this basis,a greedy strategy is adopted to identify all sub-communities for forming RPC and APC partitioned results based on the improved modularity increment,respectively.Then,the partition results are further selected based on the voltage regulation requirements.The priority regulation nodes set is further selected for voltage regulation dispatching.(2)Because of the complex solution space and implicit optimization objective function in the active and reactive power dispatching model based on voltage regulation,a high-performance Crisscross Search algorithm(CSO algorithm)is used to dispatch the active and reactive power output of distributed PVs.For RPC or APC sub-communities that require voltage regulation,CSO algorithm minimizes the reverse power flow of the point of common coupling in distribution network with high penetration PVs.The CSO algorithm has excellent search capabilities,which can perform a wide range of search in a complex solution space,and has a fast convergence speed.(3)The double-layer partition voltage regulation strategy composed of RPC and APC dispatching node selection and CSO algorithm optimization is built.Its effectiveness has been verified on the IEEE33-node and the modified IEEE123-node.After the implementation of the voltage regulation strategy based on the CSO algorithm,the voltage of each node of the distribution network is maintained the normal range,and the amount of active and reactive power that needs to be regulated for power flow dispatching is less.However,the slow optimization speed of the CSO algorithm exists in the voltage regulation dispatching model.(4)In order to improve the optimization speed of CSO algorithm,the TL algorithm is used to optimize the CSO algorithm.Based on the original CSO algorithm,this paper proposes a CSO optimization algorithm based on Transfer Learning(TLCSO algorithm).Since the daily changes of PVs have a similarity,TLCSO can greatly improve the real-time solution speed of active and reactive power dispatching when the algorithm dispatches similar data by reusing the pre-learned knowledge,which has self-learning ability.The simulation results on the modified IEEE34-node and the modified IEEE123-node all show that compared with the original CSO algorithm,this algorithm has a faster solution speed and high practicability on the basis of ensuring high solution quality.