Design of Energy Storage Photovoltaic Power Generation Device and Neural Network Method for Photovoltaic Power Prediction

Abstract

The series-parallel combination of the photovoltaic array can meet the needs of different applications such as high power or low power with the continuous optimization of photovoltaic cell materials and the increasing improvement on efficiency of photovoltaic cell. This exploration takes the independent photovoltaic power generation (PPG) device as the research goal. The system includes a photovoltaic array, Boost rectifier, single-phase photovoltaic inverter, battery and bi-directional DC/DC converter. Single-phase photovoltaic inverter is employed for the AC load power supply, and a lead-acid battery is adopted for energy storage. Buck mode and Boost mode are modeled respectively based on the analysis of the converter topology, and the designed model is employed to control the converter. Back propagation neural network (BPNN) is introduced to analyze the factors which can affect the output power of PPG. In experiment, the effective value of output voltage, output voltage, and current waveform of single-phase inverter are analyzed. The analysis results show that the output voltage changes suddenly with the load, and the device can be followed again within a few milliseconds, which means that the device has good dynamic performance. The corresponding output frequency fluctuates at a small amplitude (±1%) when the output voltage is above 220 V. Meanwhile, the distortion of the output waveform is less than 1%. The ripple coefficient of the output voltage is less than 1% when the operating mode of PPG (boost and Buck) is changed. The neural network is introduced to select the appropriate parameters (irradiation intensity, relative humidity, ambient temperature and wind speed), and the results show that the environmental temperature exerts a great influence on the output power of PPG.