Investigation of active learning techniques for dynamic Time-of-Use (dToU) tariff policy design for residential users

Abstract

The active learning approach is a special case of semi-supervised machine learning which is able to interactively query the user to reduce the uncertainty of the machine learning model. The approach is useful to minimize the data labeling cost. The project aims to study and use this method to characterize residential electricity users’ demand response to improve the prediction accuracy of energy demand. During the trials, the policies which penalize or incentivize the users to change their behaviour involve a cost associated with the grid management. Therefore, the experiments which include the above-mentioned policies are considered as cost bearing experiments. The goal of this project is to study the effect of selective sampling and random sampling of such cost bearing experiments on energy consumption prediction accuracy in simulated residential energy consumption environment. Firstly,we show the simulator design for simulating demand response of users under dynamic tariff policy.Then, we investigate two selective sampling methods- variance reduction and novelty detection. The performance of these methods under various criteria is analyzed