Efficient load management and system stability in active distribution networks (ADNs) depend heavily on accurate load forecasting and dynamic decision-making. This paper proposes a novel end-to-end joint learning model for dynamic prediction-decision joint-making in active distribution networks. The model integrates a prediction module, leveraging long short-term memory networks to forecast photovoltaic (PV) generation, active power load, and reactive power load, with a control decision module based on Twin Delayed Deep Deterministic Policy Gradient (TD3). The forecasting outputs are fed into the decision-making module, enabling dynamic adjustments to network parameters. A feedback mechanism is introduced that incorporates decision losses into the prediction module to enhance forecasting accuracy, thereby creating a tightly coupled, end-to-end joint optimization framework. In addition, a mixed precision quantization technology is adopted to reduce computational overhead and facilitate efficient deployment. Experimental results on distribution networks demonstrate that the proposed model outperforms other methods in terms of load forecasting accuracy and dynamic control effectiveness, resulting in substantial improvements in the operational efficiency and stability of ADNs.

Conventional optical imaging systems employ lenses as primary components for image formation through refracting and focusing light. However, their performance is compromised by environmental perturbations, limiting fine-detail recovery, and degrading imaging fidelity. This paper proposes deformable image learning network (DILNet), a lightweight deep deformable convolutional network for lensless imaging that addresses the constraints outlined above. Leveraging a primal-dual optimization framework, DILNet integrates a multiscale frequency-domain attention inversion mechanism, deep deformable convolutions, and a multikernel learnable point-spread function, fusing physical modeling with deep feature learning to progressively enhance detail restoration and structural integrity. In addition, DILNet unifies an AU-Net-based denoising module to capture long-range pixel dependencies, thus further improving the reconstruction quality. Comprehensive comparative and ablation studies on DiffuserCam Mirflickr dataset and PhlatCam dataset demonstrate that DILNet outperforms alternatives in mean squared error, peak signal-to-noise ratio, multiscale structural similarity, and learned perceptual image patch similarity metrics while retaining robust, high-fidelity reconstructions in demanding imaging scenarios. Combined results position DILNet as an efficient and resilient solution with strong real-world potential for lensless image reconstruction.

Rotary kiln shell temperature monitoring is essential for metallic shell protection and refractory lining maintenance in high-temperature industrial processes, while smoke, dust, thermal diffusion and non-kiln heat sources make valid shell temperature extraction difficult. This study develops a multi-scale cascaded network with low-resolution space-to-depth downsampling (MSC-LSTD) for infrared kiln shell segmentation and temperature recognition. Global infrared thermal images and local laser temperature measurements are used to construct a calibrated rotary kiln infrared dataset, and predicted kiln shell masks are mapped to temperature matrices for valid shell temperature analysis. MSC-LSTD achieves 99.82% aAcc, 99.14% mAcc and 97.03% mIoU on the rotary kiln infrared dataset, showing robust segmentation performance under complex thermal interference. The proposed framework provides a practical image-based solution for kiln shell overheating warning and refractory lining degradation assessment.