Research On Artificial Neural Network In Harmonic Suppression Of Microgrid

Since the 21 st century,the demand for electricity has been increasing.Non-renewable energy sources such as oil,coal and natural gas have become increasingly depleted,and distributed renewable clean energy such as solar energy and wind energy have developed rapidly.The distributed power supply,energy storage system and load together constitute the small power distribution system-Micro-grid.It is a new grid structure with broad prospect in the future.At present,micro-grid research is not yet mature.A large number of distributed power through the converter and harmonic generation.Micro-grid power quality produced a great impact.In order to ensure the safe and economical operation of the micro-grid,there are three kinds of usual harmonic suppression method:(1)Transforming the harmonic structure and cutting off the harmonic pollution sources;(2)Improving the electrical equipment and the anti-interference of harmonic receptor;(3)Adding harmonic compensation device transmission,and suppressing the harmonic.In this paper,the active power filter is used to suppress the harmonics of the micro grid.First of all,the paper introduces the two main types of distributed power:fed wind power generation system and photovoltaic power generation system,and establishes the model in Matlab/Simulink simulation platform.Then,several harmonic detection methods of active power filter are analyzed in detail.A adaptive neural network is designed with an improved LMS algorithm to detect the micro harmonic current.Then comparing and analysing with the traditional LMS algorithm,the results show that the improved algorithm has the advantages of rapidity,accuracy etc.Finally,the proposed algorithm combines the neural network with the PID algorithm to control the active power filter in the harmonic current tracking.The training algorithm and implementation process of CMAC neural network and PID composite controller are analyzed,and the simulation analysis is carried out in several cases.The results show that the controller can track the current value of the instruction quickly and accurately,and the result is satisfactory.