| With the promotion of a "double carbon" policy,clean and efficient distributed energy sources such as photovoltaic and wind power have been rapidly developed and applied.The lowvoltage DC distribution network connected by distributed power sources has the advantages of improving the utilization rate of renewable energy and efficiently accepting energy storage systems and DC loads,which makes it an important part of the future green and intelligent distribution system.At present,the fault detection and protection technologies related to DC systems and DC islanding detection technologies are still in the initial stage,which becomes one of the reasons limiting the large-scale application of DC networks.Loading islanding detection devices in distributed generation units within low-voltage distribution systems can effectively avoid stability and protection problems caused by the unconscious islanding phenomenon and ensure its safe and reliable operation.However,there are no electrical quantities such as frequency and phase in DC systems,and voltage as the only available islanding characteristic quantity makes the study of DC islanding detection technology more difficult.In this paper,we take the low-voltage DC distribution network as the research obj ect and conduct an in-depth study on its islanding detection technology,which is mainly as follows:The typical low-voltage DC distribution network architecture and its control strategy are analyzed,and a simplified DC distribution system model for studying islanding detection technology is built.The topologies of the on-grid photovoltaic system,energy storage system,and grid-connected units and their converter control strategies are analyzed in detail,and a simulation model is built to verify the effectiveness of the proposed DC distribution network and the control strategies of its components,laying the foundation for the study of DC islanding detection technology.A two-stage DC islanding detection method based on Maximum Power Point Tracking(MPPT)control is proposed to address the problems of the active islanding detection method such as simultaneous injection of disturbances and impact on grid power quality.The method determines the suspected islanding by detecting the DC-side current of the grid-connected converter and then uses the suspected islanding signal to trigger the injection of voltage disturbance to make the voltage of the grid-connected network fluctuate at a specific frequency after the islanding.In addition,combining the distributed power supply DC/DC converter topology and its control strategy,a small-signal model is established to analyze the frequencydomain characteristics of the parallel network voltage on the disturbance component and the optimal frequency of the disturbance signal is selected.Finally,the effectiveness of the proposed method is verified by simulation and experiments under single and multi-machine conditions.In view of the problems of traditional passive island detection methods such as large detection blind areas and difficult threshold adjustment,an intelligent passive island detection method based on the Adaptive Boosting(Adaboost)algorithm is proposed in combination with data mining technology.The method uses the feature filtering technique to obtain the important islanding feature quantity,and then relies on the Adaboost classification algorithm to generate a strong islanding classification model to achieve real-time detection and classification of DC distribution network operation status,which has the advantages of intelligent adjustment of threshold,no detection blind area and high accuracy rate.Finally,the proposed method is simulated and experimentally validated by using the constructed DC distribution network model.The results show that the method can quickly and accurately detect the islanding status of the DC distribution network. |
PDF Full Text Request