This letter proposes a novel deep learning-based multi-task approach for non-intrusive monitoring of home appliances—the first of its kind—where a network can simultaneously estimate the states and disaggregate energies of multiple appliances. An attention-powered encoder-decoder network, comprising a convolutional layer and a long short-term memory, is deployed for the above tasks. Test results from two real-world datasets demonstrate the approach’s feasibility, showcasing superior performance and reduced memory requirements.

中文翻译：

---

用于多设备非侵入式负载监控的多任务深度学习方法

这封信提出了一种新颖的基于深度学习的多任务方法，用于对家用电器进行非侵入式监控，这是同类中的第一个，其中网络可以同时估计多个电​​器的状态并分解能量。为上述任务部署了由卷积层和长短期记忆组成的注意力驱动的编码器-解码器网络。两个真实数据集的测试结果证明了该方法的可行性，展示了卓越的性能和降低的内存需求。