This paper proposes an efficient and lightweight forest smoke detection model based on YOLOv8. The model improves detection accuracy and reduces computational load by integrating a weighted feature fusion network in the YOLOv8 neck, introducing a simple and parameter-free attention mechanism (SimAM) to suppress redundant information, and applying focal modulation to enhance detection of hard-to-detect smoke and improve model speed. The improved model achieves a mean average precision (mAP) of 90.1%, which is 3% higher than the original YOLOv8. The model's parameter count and computational complexity are reduced by 30.07% and 10.49%, respectively. The model outperforms other mainstream models on a self-built forest smoke detection dataset and shows great potential for practical applications. The model is designed for edge devices with limited computational resources, enabling real-time object detection. The model's performance is evaluated using mAP, precision, and recall metrics. The results show that the improved model significantly improves detection accuracy and reduces computational complexity compared to other models. The model is also tested on different datasets, demonstrating its effectiveness in complex forest environments. The study highlights the importance of balancing accuracy and speed in forest smoke detection and suggests future research directions for improving model performance and adaptability to real-world scenarios.This paper proposes an efficient and lightweight forest smoke detection model based on YOLOv8. The model improves detection accuracy and reduces computational load by integrating a weighted feature fusion network in the YOLOv8 neck, introducing a simple and parameter-free attention mechanism (SimAM) to suppress redundant information, and applying focal modulation to enhance detection of hard-to-detect smoke and improve model speed. The improved model achieves a mean average precision (mAP) of 90.1%, which is 3% higher than the original YOLOv8. The model's parameter count and computational complexity are reduced by 30.07% and 10.49%, respectively. The model outperforms other mainstream models on a self-built forest smoke detection dataset and shows great potential for practical applications. The model is designed for edge devices with limited computational resources, enabling real-time object detection. The model's performance is evaluated using mAP, precision, and recall metrics. The results show that the improved model significantly improves detection accuracy and reduces computational complexity compared to other models. The model is also tested on different datasets, demonstrating its effectiveness in complex forest environments. The study highlights the importance of balancing accuracy and speed in forest smoke detection and suggests future research directions for improving model performance and adaptability to real-world scenarios.