研究目的
To resolve the problem of constant change in the maximum power point of photovoltaic power generation due to external temperature and radiation changes by proposing a novel method based on the self-encoding neural network technology.
研究成果
The self-coding neural network method can track the maximum power point of photovoltaic power more quickly and accurately than the conventional conductance increment method and improve the efficiency of photovoltaic power generation. It optimizes the disadvantages of complex control method and slow tracking of traditional methods.
研究不足
The method's performance under extreme environmental conditions and its scalability to larger photovoltaic systems are not discussed.
1：Experimental Design and Method Selection:
The method is based on the deep learning network training of stacking encoders and fine-tuning the self-coding neural network which utilizes the reverse propagation method with supervised learning.
2：Sample Selection and Data Sources:
Input temperature, light intensity and output D are collected online. They are data of illumination radiation from 1000W/m2 to 100W/m2 and external temperature from 25°C to 75°C, respectively.
3：List of Experimental Equipment and Materials:
MATLAB/simulink environment is used for model analysis of photovoltaic power generation system.
4：Experimental Procedures and Operational Workflow:
The data set is divided into 3501 training samples, 500 validation samples and 500 test samples. Then use data sets to train ANN.
5：Data Analysis Methods:
BP method and gradient-based optimization technology are used to change the parameters of the whole network in a top-down manner.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
