To tackle the distinctive challenges of satellite mobility—challenges that exert significant impacts on both the reliability of communication links and the efficiency of wireless power transfer (WPT)—we propose a distributed satellite-fusion cell-free (SFcf) integrated sensing and communication (ISAC) architecture. Traditional adaptive beamforming, which performs reactive compensation based on estimated channel state information (CSI), faces inherent limitations in achieving robust data transmission and efficient energy delivery within such high-dynamic environments, primarily due to its intrinsic latency. To surmount this bottleneck, we put forward a novel closed-loop ISAC framework that facilitates a proactive paradigm for the joint transfer of wireless information and power. The core concept lies in estimating the physical source of channel variation—namely the satellite’s motion parameters—and utilizing these parameters to enable predictive adaptation of both communication and energy beams. In this framework, sensing-derived Doppler and time of arrival (ToA) parameters are fed back to actively pre-compensate for channel mismatch, ensuring communication signals are robust and that energy beams remain precisely focused on the moving target. Therefore, our proposed SFcf-ISAC framework is specifically designed to validate this effective synergy, enabling sustainable 6G satellite-ground connectivity through high-efficiency WPT, remote sensing and resilient high-mobility communication.

The Deep Learning (DL) methods can automatically extract features without artificial prior, provides an effective solution for multi-temporal crop classification. Convolutional Neural Networks (CNNs) have superb spatial-spectral feature extraction capabilities, but often lack consideration of the temporal relationship of multi-temporal images, while Recurrent Neural Networks (RNNs) can better learn the sequential variation pattern. This paper designed a deep spatial-temporal-spectral feature learning network (CropNet) combining the advantages of a deep spatial-spectral feature learning module and a deep temporal-spectral feature learning module for better feature extraction in crop classification from time-series remote sensing images. From the results, the proposed method has better crop classification effects from time-series multispectral images in our experimental areas compared with some common traditional machine learning approaches and common DL methods.