Driving And Fault Detection Of Photovoltaic Inverter Based On Markov Theory

Because of the deepening of the contradiction between environmental problems and economic development,and the contradiction between limited resources and unlimited development,the thirst for new clean energy is increasing.Solarenergy has been paid attention all over the world because of its universal,harmless,huge and long-lasting advantages.The photovoltaic inverter,as the "brain" of the photovoltaic power generation system,not only shoulders the responsibilities of DC-AC,but also is the carrier of our understanding of the whole system.Accordingly,the two research hotspots came into being: driving and fault detection of PV inverter.Around these two aspects,we take the three level neutral-point-clamped(NPC)inverter as the research object.Then elaborate the core theoretical basis——Markov theory,including the Markov chain,the hidden Markov model(HMM)and the related algorithms.At the same time,we use genetic algorithm to improve the defect of HMM,and give the steps of combining HMM with genetic algorithm.The specific research content is divided into the following two aspects:(1)For the driving mode of the photovoltaic inverter,we expound the principle of the traditional SVPWM drive technology.In view of its fixed switching mode,the high amplitude harmonics are concentrated at the integer times of the switching frequency to form the defects of the electromagnetic interference.We propose to generate random factors through the Markov chain,and use this random factor to randomize the original switching frequency and the action time of small vector in the traditional SVPWM technology.Therefore a hybrid random SVPWM driving method based on Makrov chain is formed.Finally,through the MATLAB/Simulink simulation experiment,we verify that the driving method not only greatly reduces the harmonic amplitude that concentrates on the integer times of the switching frequency,makes the spectrum more continuous,but also improves the shortcoming of the random number without Markov chain which has less randomness and uneven distribution.(2)For the fault detection of photovoltaic inverter,we first analyze the type of fault,and elaborate the commonly used detection methods.In view of the shortcomings of these methods,such as poor practicality,low efficiency and low detection accuracy,we combine HMM with genetic algorithm to form a new model——GHMM.Then we train the GHMM in each fault.Finally,these models are used for fault detection.We construct the experimental platform,implement the related algorithms through MATLAB and C language programming and complete the fault model training and fault detection.Comparing with other methods such as the unimproved HMM,BP neural network and support vector machine,the training time of the fault model is greatly reduced and the recognition accuracy is increased by 22.2%,reaching 98.3%.The superiority of using GHMM in fault detection of inverter is verified.