Control of a photovoltaic source emulator using artificial neural network

Abstract

The photovoltaic (PV) emulator is a nonlinear power supply that produces a similar current-voltage characteristic of the PV module. However, the PV emulator output is volatile due to the nonlinear characteristic of the PV module. Conventionally, the overdamped PV emulator is required to prevent instability but results in slow dynamic response. On the other hand, the dynamic response of the PV emulator varies with changes in solar irradiance, ambient temperature and output resistance. The researches carried out in recent years for the control techniques include direct calculation method, look-up table method, piecewise linear method, neural network method, and curve segmentation method. Each of the method has advantages and disadvantages in terms of processing burden, memory required, accuracy, adaptability and independency. This research project focuses on the simulation of a combination of interleaved buck converter with two-stage inductor and capacitor filter to improve the dynamic performance of the PV emulator. Artificial neural network is used to overcome the complexity in the adaptive proportional-integral (PI) controller to achieve a stable and fast dynamic response of the PV emulator. The proposed control technique is simulated using MATLAB/Simulink® simulation package with varied output resistance and irradiance. ANFIS Editor toolbox is used for the training and learning process. The PI gains of the conventional method are set to limit output current overshoot under various output resistance. By comparison to conventional method during start-up response, the proposed control technique shows improvement of 40% to 90% faster in dynamic performance of the output current.