Optimal Control Technology Of Active Distribution Network Based On Physical-data Fusion Model

In recent years,the large-scale access of distributed generation has not only alleviated the energy shortage,reduced the power supply pressure,improved the power supply efficiency and increased the operational flexibility of the power system,but also brought new problems and challenges to the distribution network.The distributed generation supply with high proportion of access changes the operation characteristics of distribution network,affects the safety,reliability and economic operation of distribution network,and brings difficult problems to the optimization and regulation of distribution network.Based on the actual operation characteristics of the distribution network and considering the status and demands of regulation and control at different time scales,this paper studies the optimization regulation technology of the distribution network and carries out the following work:1)The high proportion of access to distributed generation has changed the structural characteristics and operation characteristics of the distribution network.With the development of information technology and artificial intelligence,new ideas have been provided for the optimization and control of the distribution network.Based on this,the basic model of the optimization and control of the distribution network has been studied to lay a foundation for the optimization and control of the distribution network.Firstly,the probability information of distributed photovoltaic and load uncertainty is modeled,and the probability model of its parameter distribution is established.Secondly,the distribution network system is modeled to establish its power line model and power flow calculation model.Finally,the distribution network measurement database is modeled,the source of measurement data is analyzed and data fusion is carried out.The basic model is established from the perspectives of node,system and data,which lays the foundation for the optimization and regulation of active distribution network based on physical-data fusion model.2)The optimization and control of distribution network on a long time scale is generally carried out on the basis of prediction information.However,different time scales have different optimization requirements and equipment response characteristics.Therefore,the day-ahead and intra-day optimization of distribution network is studied based on data-guided and physical-driven method.Firstly,considering the response characteristics of discrete and continuous regulation equipment,the optimal dispatching framework of distribution network is designed.Secondly,the deterministic voltage near violations risk constraint is designed in the day-ahead phase,a probabilistic risk assessment method of uncertain voltage is proposed in the intra-day phase,which quantitatively evaluates the voltage operation state of distribution network through a small number of input random variable samples and deterministic calculation.Finally,according to the characteristics of the objective function and the power flow model,the second-order cone relaxation is carried out on the equation of the square term in the Distflow power flow equation,and the parameters of the objective function and the solution form are set reasonably.On the basis of guaranteeing the economic operation of distribution network,the risk of node voltage safety is eliminated.3)Online optimization of distribution network is generally carried out on the basis of measurement information,but there are few measuring devices and insufficient online measurement information in the distribution network.Therefore,research on online optimization of distribution network is carried out based on physical-guided and data-driven method.Firstly,the online model of distribution network is established through generative adversarial networks,and the operation data of distribution network can be generated according to the historical objective law through the trained generative adversarial networks.Secondly,the generator input is optimized by reasonably designing the objective function to complete the missing data of the online measurement of the distribution network and realize the system state reconstruction.Finally,the objective function and constraint conditions are further designed to optimize the generator input and complete the online optimization of the distribution network.The online optimization of distribution network is realized only through partial node measurement data without relying on power flow calculation model.