Identification of the Potential of Residential Demand Response Using Artificial Neural Networks

The uncertain and volatile nature of renewable energy tends to cause imbalance in the supply and demand of electricity. Residential demand response has emerged as a methodology to alleviate such imbalances in the electricity grid. By utilizing the flexibility of the electricity demand, residential demand response programs offer monetary incentives to the customers to bring about temporary reductions in energy consumption when the grid faces supply shortage or is overloaded. Utilities are interested in enrolling residential customers in their demand response programs in order to maintain a healthy demand-supply balance. Selecting the right set of customers for this is a demanding task, and what makes it even more challenging is the lack of use of the temporal consumption data in the existing customer recruiting methodologies. In this paper, by utilizing the granular smart meter data, we develop a methodology that estimates a customer's potential reduction in consumption based on his/her sensitivity to external temperature, which is then used to train a mixture density recurrent neural network. The network outputs the probability distribution of the demand reduction of the customer. The results can be utilized to accurately select customers, which can then ensure the reliable performance of the demand response program.