Automated Demand Response Contracts when Partial Consumption Information is Available

Automated demand response (ADR) programs play a key role in alleviating the energy supply and demand mismatches in a pre-contracted way. In this paper, we focus on the design of effective incentive-based ADR contracts by energy providers and retailers when partial information only is available. We assume that the consumption of a small subset of users is monitored closely, by means of appliance-level smart meters. The provider thus acquires detailed knowledge of their energy profiles and preferences, which is subsequently used to extract information for the rest of the users. We rely on the assumption that users with similar characteristics would behave in a similar way in terms of energy consumption; hence they could be characterized by the same utility function and be offered the same ADR contract. We employ clustering and classification techniques in order to group users, and then we use a specific representative utility function for each appliance for all users of a group. We then apply our previously published ADR incentives' calculation algorithm together with a policy restricting the user discomfort. Thus, our contribution is a complete and practically applicable approach for ADR, which according to our experimental evaluation results is effective despite the inevitable inaccuracy due to clustering. Our results also indicate that it is beneficial for the energy provider to offer a few choices of ADR contracts to each user.