一种利用LSTM和稳态时间序列的光伏阵列故障诊断方法

受最大功率点跟踪算法和时变环境条件的影响，光伏阵列的电气工作参数包含了复杂的暂态过程以及工频干扰噪声，严重影响了故障特征质量以及诊断算法性能。针对该问题，本文首先提出了一种基于最大功率点（MPP）的稳态时间序列预处理方法，以自动过滤数据中的暂态过程和干扰噪声，获取连续的稳态时间序列电气特征数据，作为故障诊断模型的输入参数；然后，提出了一种基于长短期记忆网络（LSTM）的深度网络模型，以实现对光伏阵列常见故障的检测及分类；最后，在一个小型光伏并网发电系统及其Simulink仿真模型上，进行故障模拟及仿真以验证所提出的故障诊断方法。实验结果表明所提出的故障诊断方法具有良好的精度和泛化性能，并且优于常规的反向传播神经网络（BPNN）和循环神经网络（RNN）。

A photovoltaic array fault diagnosis method using LSTM and steady-state time series

Affected by the maximum power point tracking algorithm and time-varying environmental conditions, the electrical operating parameters of the photovoltaic array include complex transient processes and power frequency interference noise, which seriously affect the quality of fault characteristics and the performance of the diagnostic algorithm. In response to this problem, this paper first proposes a steady-state time series preprocessing method based on the maximum power point (MPP) to automatically filter the transient process and interference noise in the data to obtain continuous steady-state time series electrical characteristic data. As the input parameters of the fault diagnosis model; then, a deep network model based on long short-term memory network (LSTM) is proposed to realize the detection and classification of common faults in photovoltaic arrays; finally, in a small photovoltaic grid-connected power generation system and On the Simulink simulation model, fault simulation and simulation are performed to verify the proposed fault diagnosis method. The experimental results show that the proposed fault diagnosis method has good accuracy and generalization performance, and is better than the conventional back propagation neural network (BPNN) and recurrent neural network (RNN).