Long-term time series prediction is a challenging and essential task both in theory and practice. Recently, information granulation is shown to be an appropriate tool for the long-term forecast. Though some models for the long-term prediction problem have been proposed using information granulation recently, there is still a growing need to develop new prediction approaches for time series data based on information granule, which can capture the dynamic trend change with high accuracy. In this paper, a long-term prediction approach, based on back-propagation neural network and information granule, is proposed. Firstly, the individual numerical intervals for the time series are obtained by using the principle of justifiable granularity in information granule. Then, an automatic linear trend extraction method is developed to extract the trend change, which is inherited in granules. Finally, a hierarchy of neural network is constructed to carry out prediction by using information granule as input. Experiments using publicly available time series datasets demonstrate that the proposed approach can achieve better performance than the existing models for long-term prediction.

With the popularity of photovoltaic power generation, the number of photovoltaic power plants is gradually increasing, and the sustainability, efficiency, stability and safety of photovoltaic power plants have become particularly important. For this reason, it is necessary to regularly detect and diagnose photovoltaic modules, and to replace defective photovoltaic modules in a timely manner. In this work, a new image classification network based on the MPViT network structure is designed to solve the problem of fault detection and diagnosis of photovoltaic panels using image processing methods. This model is based on the MPViT model, aiming at the similarity between different types of photovoltaic panel images, the Enhanced Local Self-Attention block (ELSA block) is introduced to enhance the learning ability of the model's local detail features, so as to improve the model's ability to discriminate categories. In this paper, binary classification and multi-classification experiments are carried out on two datasets, one is a photovoltaic module dataset based on infrared imaging (dataset1), and the other is a photovoltaic module dataset based on electroluminescence imaging (dataset2). The M-E model (MPViT model introduced with ELSA block) achieves 94.1% and 88.5%, 90.7% and 86.4% accuracy for binary classification and multi-class classification on dataset1 and dataset2 respectively, better than MPViT.