Demand response model identification and behavior forecast with OptNet: a gradient-based approach

Published in the Thirteenth ACM International Conference on Future Energy Systems, 2022

Recommended citation: [1] Y. Bian, N. Zheng, Y. Zheng, B. Xu, and Y. Shi, “Demand response model identification and behavior forecast with OptNet: a gradient-based approach,” in Proceedings of the Thirteenth ACM International Conference on Future Energy Systems, Virtual Event, Jun. 2022, pp. 418–429. doi: 10.1145/3538637.3538871. https://dl.acm.org/doi/abs/10.1145/3538637.3538871

Price-responsive demand side resources can adjust their energy usage in response to time-varying price signals, which provide flexibility and promotes system reliability. In this work, we propose a novel data-driven approach that incorporates prior model knowledge for predicting the behaviors of price-responsive demand resources. We propose a gradient-descent method to find the model parameters given the historical price signals and observations. We prove that the identified parameters will converge to the true user parameters under a class of quadratic objective and linear equality constrained demand response (DR) models. We demonstrate the effectiveness of our approach through numerical experiments with synthetic data using demand models including batteries, buildings, and aggregations of price-responsive loads. The proposed approach significantly improves the accuracy of both DR model identification and behavior forecasting compared to previous black-box data-driven approaches and inverse optimization approaches.

Download paper here

Recommended citation: [1] Y. Bian, N. Zheng, Y. Zheng, B. Xu, and Y. Shi, “Demand response model identification and behavior forecast with OptNet: a gradient-based approach,” in Proceedings of the Thirteenth ACM International Conference on Future Energy Systems, Virtual Event, Jun. 2022, pp. 418–429. doi: 10.1145/3538637.3538871.