FFYOLO is a lightweight forest fire detection model based on YOLOv8, designed to improve detection accuracy and efficiency. The model introduces several key components: a Channel Prior Dilatation Attention (CPDA) module to enhance feature extraction, a Mixed-Classification Detection Head (MCDH) for improved classification, and MPDIoU loss to enhance regression and classification accuracy. Additionally, a lightweight GSConv module is used to reduce parameters while maintaining model accuracy, and knowledge distillation is applied during training to improve generalization and reduce false detection. Experimental results show that FFYOLO achieves an mAP0.5 of 88.8% on a custom dataset, which is 3.4% higher than the original YOLOv8 model, with 25.3% fewer parameters and 9.3% higher FPS. The model is optimized for deployment on low-power devices and performs well in complex forest fire scenarios. The paper also discusses the challenges of forest fire detection, including the need for accurate and efficient detection in diverse environments. FFYOLO addresses these challenges by combining advanced attention mechanisms, improved detection heads, and efficient convolutional modules, making it a promising solution for real-time forest fire detection.FFYOLO is a lightweight forest fire detection model based on YOLOv8, designed to improve detection accuracy and efficiency. The model introduces several key components: a Channel Prior Dilatation Attention (CPDA) module to enhance feature extraction, a Mixed-Classification Detection Head (MCDH) for improved classification, and MPDIoU loss to enhance regression and classification accuracy. Additionally, a lightweight GSConv module is used to reduce parameters while maintaining model accuracy, and knowledge distillation is applied during training to improve generalization and reduce false detection. Experimental results show that FFYOLO achieves an mAP0.5 of 88.8% on a custom dataset, which is 3.4% higher than the original YOLOv8 model, with 25.3% fewer parameters and 9.3% higher FPS. The model is optimized for deployment on low-power devices and performs well in complex forest fire scenarios. The paper also discusses the challenges of forest fire detection, including the need for accurate and efficient detection in diverse environments. FFYOLO addresses these challenges by combining advanced attention mechanisms, improved detection heads, and efficient convolutional modules, making it a promising solution for real-time forest fire detection.