Residential Load Modeling and Forecasting

The expansion of residential demand side management programs and increased deployment of controllable loads is changing the way electricity is consumed in residential applications. Understanding the characteristics of these loads is essential to support efficient grid operation and to optimize home energy consumption for the consumer. Short-term or near real-time load forecasting at the home or appliance level is important for the provision of various services to the grid and the consumers, including demand-side management and home energy management systems. However, modeling and forecasting of demand at the appliance level is challenging due to the intrinsic variability and uncertainty of exogenous environmental variables and human behavior.

In this paper, we develop a statistical load model which describes the probabilistic nature of residential appliance demand and a scalable learning method for short-term load forecasting. In particular, we develop a conditional hidden semi-Markov model (CHSMM), which is based on two characteristics observed from appliance load dynamics: (1) discrete appliance operating states and (2) random time durations spent in each state. The state transition probabilities and emission and duration distributions of the CHSMM are conditioned on exogenous variables, which allows factors such as temperature and seasonal effects that impact the physical dynamics of the load to be incorporated into the model. Case studies performed using granular sub-metered power measurements from the Pecan Street database for various types of appliances demonstrate the effectiveness of the proposed load model for short-term prediction.

Publications

Data-Driven Load Modeling and Forecasting of Residential Appliances
Yuting Ji, Elizabeth Buechler, and Ram Rajagopal
IEEE Transactions on Smart Grid (2020)

Presentations

Data-Driven Load Modeling and Forecasting of Residential Appliances
Elizabeth Buechler, Yuting Ji, and Ram Rajagopal
Presentation at the 2021 IEEE PES General Meeting (virtual)