Application of Sample Convolution and Interaction Network to Time Series Prediction Based on Power Line Carrier Communication

Power line carrier communication is a technology that uses existing low-voltage distribution networks as transmission media to achieve data transmission and information exchange. However, with the access of photovoltaic subsystems, their prediction and debugging of time series signals become more difficult. In this paper, in view of the errors in time series prediction, sample convolution and interaction network (SCINet) is used to fix the problem, and it is an innovative multi-layer neural network framework based on the characteristics of time series. Experimental results show that SCINet achieves significant forecasting accuracy improvements over both existing convolutional models and Transformer-based solutions across electric series forecasting datasets.