Household electricity load forecasting using historical smart meter data with clustering and classification techniques

The uptake of smart grid technologies and increasing deployment of smart meters have brought greater attention on the analysis of individual household electricity consumption. Within the smart grid framework, home and battery energy management systems are becoming important demand side management tools with various benefits to households, utilities and networks. Load forecasting is a vital component of these tools, as it can be used in optimizing the schedule of household appliances and energy operations. Inspired by the advancements in larger scale load forecasting, this paper proposes a novel forecast method for individual household electricity loads. Besides using smart meter data together with weather and temporal variables, which are commonly used in more conventional household load forecasting methods, this approach integrates the information contained in typical daily consumption profiles extracted by clustering and classification methods. In addition to the improvements in forecast performance, the method reveals key information about a household’s habitual load profiles and other important variables which impact household consumption.