基于脉冲神经网络光伏电站热斑故障预测

Hot spot fault prediction of photovoltaic power station based on pulse neural network

针对光伏电池的热斑现象,利用红外热成像仪对光伏电池进行实时扫描得到红外热图,对红外热图进行图像增强等预处理。将图像分为四种状态,转换为信息量较强的状态矩阵。搭建基于脉冲时间依赖的突触可塑性(STDP)算法的带泄漏整合发放(LIF)模型脉冲神经网络,以发生典型热斑故障的电池片的状态变化作为时间序列训练脉冲神经网络,使模型习得热斑故障的状态时序,从而检测热斑故障,发出警报并实现定位的功能。脉冲神经网络的训练难度较大,因此,在Matlab上搭建了目前广泛使用的反向传播(BP)神经网络模型,通过对模型的改进,也能实现检测热斑故障的效果,继而证明了脉冲神经网络检测热斑故障的可行性。

Aiming at the hot spot phenomenon of photovoltaic cells,the infrared thermal imager was used to scan the photovoltaic cells in real time to get the infrared thermal image,and the infrared thermal image was preprocessed by image enhancement.The image was divided into four states and converted into a state matrix with strong information.The pulse neural network of LIF model based on STDP algorithm was built.The pulse neural network was trained by taking the state change of cells with typical hot spot fault as time series for making the model learn the state sequence of hot spot fault to detect hot spot fault,send out alarm and realize the positioning function.Because of the difficult training of pulse neural network,the widely used BP neural network model was built on Matlab.Through the improvement of the model,the effect of detecting hot spot fault can be achieved,proving the feasibility of detecting hot spot fault by pulse neural network.