Apply RECS Hourly Loads Databases to Develop Excel Agent-Based Energy and Hourly Loads Forecasting Models

Energy and hourly load forecasting typically requires statistical knowledge applied in econometric models or in-depth appliance engineering data and market information used in end-use models.  Agent based models apply an intuitive framework that models individual households, i.e., agent, behavior.  This approach requires neither a statistical background nor detailed appliance engineering knowledge to produce model forecasts that, in many ways, are preferable to traditional econometric and end-use models.

This paper describes the development and application of an easy-to-use agent-based Excel model that applies an extended version of The Department of Energy’s 2020 Residential Energy Consumption Database (RECS, final release, June 2023).  The RECS Database has been extended by Jackson Associates to include whole-building and end-use 8,760 hourly kW loads for each of the 18,400+ households in its national survey.  

Agent based models have been used in a variety of industry applications for decades to provide valuable insights not available with econometric and end use models.    RECS Extended Hourly Loads Databases provide a relatively simple agent-based easy-to-use Excel framework to forecast future energy demands and hourly loads.  Model forecasts reflect impacts of electric price, income and equipment efficiency increases along with various demand management initiatives and other factors on individual households. 

A short paper describing this application is available at  https://maisy.com/recs_abm.htm .  The charts show impacts of dwelling unit vintage updates (DU UPDATES), income and price impacts (YELAS and PELAS) and efficiency increases (EFFIC) on hourly loads for the first day in January and annual GWH electricity use for a single scenario analysis.  Increases in dwelling units, electricity use and August peak kW are presented in the bottom right chart.