Long non-coding RNAs (lncRNAs) are important for plant growth, how plants respond to stress, and their overall development. However, it can be difficult to identify them accurately because they come in many structures and can look similar to coding RNAs. In this study, we introduce DeepPInc V2.0, a new tool that uses deep learning to analyze both the sequence and the secondary structure of RNAs, combining them in a DenseNet-CNN hybrid model. DeepPInc V2.0 outperforms existing tools on various plant datasets, achieving an accuracy of 94.2%, an F1-score of 0.93, and a Matthews Correlation Coefficient (MCC) of 0.88. It consistently outperforms seven leading tools in this area. Importantly, the model still works well even with incomplete or shortened sequences, which is a common issue in transcriptome studies. When we applied DeepPInc V2.0 to the wheat (Triticum aestivum) transcriptome under heat stress, we found over 27,000 possible lncRNAs. Of these, 1,830 were expressed differently, suggesting they may play a role in helping plants adapt to stress. These results demonstrate that DeepPInc V2.0 is a reliable and accurate platform for identifying lncRNAs in plants, facilitating the study of large RNA collections and understanding the functions of non-coding elements.