I. Introduction
Load forecasting in power system is one of the most important key of helping the system control as well as grid planning. Especially in distribution networks, an accurate load forecasting can greatly facilitate the system control, which contributes a lot of the system stability. The accurate load forecasting also helps the power system to be aware of the flow distribution across the grid, for example, varying of the load, valley of the load, and peak of the load. However, in the current power grid, two critical factors introduce significant uncertainty and randomness of load forecasting. The first is the rapid growth of the system scale, whilst the second is the deployment of the new energy including wind power and photovoltaic power. Therefore, an accurate load forecasting approach which can reduce the errors caused by the uncertainty and randomness is anxiously needed for serving the modern power system.