| With rapid development of distributed photovoltaic power generation, it is more efficient, stable, secure, and controllable in the transmission of grid system that attracts more and more researchers to pursue. There are many factors that restrict power transmission, and one of them is islanding effect which is a phenomenon that cannot be ignored. Once the islanding occurs, it will affect the switching protection action of the distribution system, endanger the security of power grid maintenance worker, and is extremely easy to damage electrical equipment when reclosing. Therefore, to prevent adverse effects of the islanding, the development of universal and effective islanding detection and protection technology is a crucial study to promote sustainable development of large-scale applications of renewable energy and new energy power systems.Currently, the situation of large scale integration of distributed generation system,passive islanding detection methods are not required to be embedded in the control strategy of distributed generation, and have good adaptability to all kinds of distributed generation systems with different topologies and control strategies. Among them,passive islanding detection methods based on signal processing technology have the characteristics of small non-detection zone, fast response, and strong adaptability to power grid disturbance. Thus it has been widely researched and developed rapidly in recent years. Therefore, this work focuses on studying various islanding detection methods based on the wavelet packet transform in grid connected photovoltaic inverter.The main research contents and innovations are as follows:(1) The variations of PCC point voltage harmonics before and after islanding occurs are analyzed. The Fast Fourier Transform, Wavelet Transform and Wavelet Packet Transform (WPT) are used to analyze the PCC point voltage harmonics and after results are compared. WPT is found to be the best in islanding detection. In addition, to solve the problem of selecting the best wavelet basis function, the MDL index of the PCC point voltage signal is calculated according to the minimum description length criterion,and then the comparative analysis shows that Daubechies db5 is the best wavelet basis function.(2) The voltage wavelet packet relative to logarithmic energy entropy value of the node (3, 4) at PCC is used as the islanding feature. Hence, an islanding detection algorithm based on wavelet packet transform (WPT) is proposed, and verified by simulation. Simulation results show that the proposed WPT algorithm can be able to detect islanding in the non-detection zone of OVP/UVP and OFP/UFP methods. In addition, it has the fast detecting performance.(3) Multi-voltage wavelet packet relative to logarithmic energy entropy at PCC is constructed and used for the design of a pattern recognition BP neural network. An islanding detection algorithm based on wavelet packet transform and BP neural network is investigated. The performances of the proposed algorithm in single- and multi-parallel inverters with different islanding and non-islanding cases are validated by simulation and experimentation. The detailed results show that the proposed algorithm is fast, effective, and accurate to detect islanding conditions. In addition, the proposed technique can be applied to both single and multi-parallel inverters.(4) An islanding detection algorithm based on the wavelet packet transform and phase shift is proposed, and its simulation and experimental verification are carried out.The verified results show that the proposed algorithm can achieve effective islanding detection with different load characteristics and quality factors. Moreover, the proposed algorithm is not affected by dilution effect in the case of multi-parallel inverters and has no non-detection zone with Qf < 4. In addition, the proposed algorithm is simple to implement and when the inverter is grid-connected it will not have an effect on the power quality. |
PDF Full Text Request