Power load forecasting is a crucial component of smart grid management, playing a pivotal role in resource optimization, energy scheduling, cost control, and sustainable development. With the development of deep learning technology, time series prediction models have shown great potential in power load forecasting. However, existing research literature focus on classical statistical models and traditional deep learning algorithms, and the research and application of novel time series prediction models are still insufficient. To address this, this paper summarizes the development and characteristics of novel time series forecasting models and discusses their latest research and applications in power load forecasting. This paper first summarizes the background and task definition of power load forecasting; then, it systematically reviews the research progress of novel time series forecasting models and selects four representative models to be validated on public datasets. Finally, it demonstrates the models’ performance and applicable scenarios through comparative analysis, providing inspiration for model selection and future research.