Online Load Model Parameter Identification For Multiple Scenarios Based On Multi-Source Data

With the expansion of the scale of modern power system and the deepening of its interconnection,the importance of power load model becomes more and more prominent.Accurate real-time load model plays a key role in online security and stability analysis,preventive control and optimal operation of power system.With the popularization of various operation monitoring and data acquisition equipment in smart grid,load operation data is constantly enriched and data quality is gradually improved.Based on this,load modeling scenarios has been divided according to the composition of available load data,and the online load model parameter identification for multiple scenarios by integrating multi-source load data has been carried out.The main works are as follows:(1)In order to meet the needs of load model identification at different voltage levels,a construction method of typical load model library is proposed.For different types of user nodes,different 380/220V equipment models are obtained by laboratory test or field measurement.For nodes with voltage levels of 10kV and above,representative load nodes are selected by clustering analysis according to their industry compositions,and then the load models of typical nodes are obtained by bottom-up aggregation.Therefore,typical load model libraries of various voltage levels can be established to provide static and dynamic basic model for online load model identification.(2)Based on the advantages of the component-based method and the measurement-based method,an online two-layer load model parameter identification method based on an aggregation-identification structure is proposed.Firstly,a node synthesis load model(SLM)considering the effect of distribution network impedance is established by the component-based method based on the sub-item data reflecting the load composition of a specific node.Then,the SLM dominant parameters are determined by perturbation method.Finally,the model dominant parameters are identified by an improved particle swarm optimization algorithm based on the real-time fault measurement data.(3)For the scenarios that the node load composition is unknown,a load component decomposition and online modeling method based on small disturbance response characteristics is proposed.The optimization model for load component decomposition based on small disturbance data is constructed and an optimization model solution method based on artificial neural network(ANN)is designed.Monte Carlo simulation algorithm is used to construct simulation data for offline ANN training,and the load components of nodes were determined dynamically base on small disturbance measurement data in online application.Based on the typical load model library,the component-based method is adopted to build the node SLM,which supported the load model dynamic update in the quasi-steady state scenarios.Different simulation cases designed on CEPRI-36 node system and the engineering case in real system have demonstrated that the proposed methods can effectively improve the efficiency and accuracy of online load model identification in fault scenarios and quasi-steady scenarios.