利用改进初始化的残差-密集连接网络的光伏阵列故障诊断

针对在现实生活中光伏阵列大部分运行在正常的工作状态,缺少故障数据的问题,提出一种改进初始化的方法代替随机初始化来训练深度学习模型,以提高故障诊断模型的可靠性.同时,提出基于残差-密集连接网络的光伏故障诊断模型,并基于I-V曲线与最大功率点、温度、辐照度和填充因子作为输入特征.最后,通过多种光伏阵列故障数据检测所提出的方...

A fault diagnosis of photovoltaic array based on improved initialization residual-dense connection network

In real life, most of photovoltaic arrays are running in normal working state, lack of fault data. In order to improve the reliability of fault diagnosis model, a new method of improving initialization instead of random initialization is proposed to train the deep learning model. At the same time, a photovoltaic fault diagnosis model based on residual dense connection network is proposed, and I-V curve, maximum power point, temperature, irradiance and fill factor are used as input features. Finally, the performance of the proposed method is tested through a variety of photovoltaic array fault data. The experimental results show that the improved initialization residual dense connection network still has high convergence speed, high accuracy and high stability in the case of small samples, and can stably classify photovoltaic array faults in various environments, which has practical significance.