基于神经网络的光伏并网逆变器分析-电力系统及其自动化专业论文.docx

II II Abstract The conventional energy sources for electrical power include hydroelectric, fossil fuels, and nuclear energy. The wide use of fossil fuels has resulted in the problem of greenhouse emissions worldwide. This also seriously damages the earths environment. Besides, fossil fuels will be exhausted in the future, and their cost has obviously increased. Studies of photovoltaic (PV) generation systems are actively being promoted in order to mitigate environment issues such as the green house effect and air pollution. The target of control in photovoltaic grid-connected system is ensuring the output current of the inverter circuits to be the sine waves of quality, with the same frequency and phase as the grid. Based on the features of grid-connected inverter devices, a kind of control tactic that combines the multi-variable double-loop control with the synchronization phase-locked control is presented. The multi-variable double-loop control is used for current real time tracing to eliminate the principle error existing in the single-loop control. The synchronization phase-locked control guarantees the synchronization of output current of inverter with system voltage. In Photovoltaic Grid-connected system, inverter is key part, whose efficiency and reliability will influence the performance of the whole system. Because of the high voltage utilization ratio and easy to realize, space vector pulse width modulation (SVPWM) has widespread application. Realize SVPWM needs fast and the massive operations, is the typical non-linear link. The Radial Basis Function ( RBF ) Neural-Network has the unique superiority in the realization function fitting aspect, the structure reasonable nerve network can approach a nonlinear function by the free precision, this provide to another way for realized SVPWM obviously. Basing on the theory on Space-Vector PWM, the paper describes the mapping implemented by Neural-Network from tracked voltage vector to turn-on time of IGBT about re