Research On Grid-Connected Islanding Detection For Inverter-Based DG

The islanding detection method studied in this paper is based on grid-connected of inverter-based DG Firstly, it needs to analysis and design every part of the grid system and build the grid-connected model in the transient simulation software PSCAD/EMTDC. The simulation simulates eleven kinds of operating status, such as normal, islanding and a variety of short circuits. Collect the three-phase current signals of each state and extract the feature in Matlab. Finally, the feature vector will be analyzed AINMC algorithmic to realize the islanding detection. The method studies in the paper are a passive islanding method which does not affect the power quality. It has a high accuracy, a small non-detection zone (NDZ) and a certain reference value of practical application.The paper includes the following sections:The opening chapter gives an introduction of the grid-connected mode and standard, analyze its impact on distribution network. Study the meaning and research status at home and abroad of the islanding detection.In the second part, the paper studies the mechanism, the standard and the non-detection zone of islanding detection. Analyze the methods, principle, advantages, disadvantages and application occasions of the existing detection. The passive detection method has no effect on power quality, but it has a large NDZ. To solve this problem, the paper explores a new passive detection method.The designed grid-connected model mainly includes three parts:inverter topology, LCL filter and the control method of the inverter. On the basis of studying the inverter topology, select three-phase voltage source inverter as the grid-connected inverter structure. Then analysis and calculation the LCL filter parameter and verify that the design of the filter has a good performance. At last, study the control mode of the grid-connected inverter, the system uses double-loop control. Design the voltage and current loop PI regulator and simulation its functions.In the forth and fifth chapter, a new islanding detection method is studied. In chapter4, the wavelet packet singular entropy was applied to extract the eigenvalues of three-phase current signals retrieved at PCC point and take it as the feature vector. The experiments show that the feature vector of every state is separate and can help effectively distinguish island status.In chapter5, the features are put into artificial immune network classifiers to realize island detection. In this process, the improved AINMC algorithm was utilized to train the feature and the KNN algorithm to classify. The whole process is implemented by programming in Matlab. The result shows that the output categories of the classifiers in line with expected result when the classified parameters (the feature vector of unknown running state) are put into the classifier. The islanding detection method studied in this paper has a good effection.