An Online Learning Method for Industrial Demand Response Based on Load Disaggregation

Industrial demand response (DR) plays an important role in the demand side management. In this paper, a DR pricing problem is formulated in which the utility offers an incentive price to industrial users and a linear response function of industrial user is considered. We propose an online learning and pricing algorithm considering load disaggregation with error to learn the industrial users’ DR potential and determine the real-time price to minimize the difference between total response and the target. Specifically, EM algorithm and random exploration are applied to estimate model parameters. Numerical simulations validate the performance of the proposed algorithm. Comparison with a baseline method without load disaggregation and a baseline method ignoring load disaggregation error demonstrates the effectiveness of our method.