Island detection method of self-adaptive grid-connected photovoltaic system based on machine learning

Abstract

The invention discloses an island detection method of a self-adaptive grid-connected photovoltaic system based on machine learning. The island detection method comprises steps that step (1):historic operation data of a photovoltaic system and simulation data of a model established according to an actual photovoltaic system are acquired, and the basic database of the three-phase voltage and current of the grid-connected operation and the island operation of the photovoltaic system under various conditions is established; step (2): the three-phase voltage and current signals of the basic database are processed, and electrical characteristic quantities are selected and extracted, and the electrical characteristic quantities and operation states are used to form a characteristic quantity database; step (3): the characteristic quantity database is analyzed by using an Adaboost algorithm, and a two-category model used to determine the island operation state and the grid-connected operation state of the photovoltaic system is established, and the accuracy of the system operation state determined by the two-category model is verified by a verification sample set; step (4): by acquiring the electrical characteristic quantities of the system in the current operation state generated in the step (2) and inputting the electrical characteristic quantities into the two-category model generated in the step (3), whether the system is in the island operation state is determined.

Description

technical field [0001] The invention relates to the technical field of power system analysis, in particular to an islanding detection method for an adaptive grid-connected photovoltaic system based on machine learning. Background technique [0002] Solar photovoltaic power generation technology is of great significance for saving conventional energy, protecting the environment, and promoting economic development. In recent years, it has been widely valued internationally and has achieved rapid development. At present, my country's photovoltaic installed capacity ranks first in the world. Photovoltaic system island detection has now become one of the key points of attention in the protection of grid-connected new energy systems. Regulations: The photovoltaic power generation system should have the ability to quickly monitor islands and immediately disconnect from the grid, and the anti-island protection action time should not exceed 2s. [0003] At present, the islanding det...

AI Technical Summary

Problems solved by technology

However, the accuracy of the judgment model generated by it is limited, and its accuracy is only the average of the accuracy of multiple inspection results under a specific sample, which cannot effectively improve the detection accuracy, so the reliability of its judgment of isolated islands is questionable

Images