Fairness in Real-Time Energy Pricing for Smart Grid Using Unsupervised Learning

Abstract

The capabilities of the Smart Grid coupled with dynamic pricing enables the Smart Grid to adaptively manage the electricity generation and distribution. Several dynamic real-time pricing schemes have been proposed in recent times but few have been successfully implemented despite their economic and environmental benefits. In particular, the current real-time pricing schemes have not been able to incentivize subscribers to respond to time-varying prices in order for the smart-grid to fully benefit from such pricing. Traditional pricing schemes failed to incorporate fairness for end-users because real-time data gathering was expensive and impractical before smart devices were incorporated into the grid. In this paper, we propose a novel dynamic pricing model, fair dynamic pricing (FDP), to maintain reliable power supply during times of peak demand. The proposed model is analyzed and evaluated using a real-time consumer load dataset from San Diego, CA, USA. We demonstrate that in periods of peak load, the burden of generating expensive electricity is placed on subscribers responsible for creating the peaks rather than being subsidized by the remaining subscribers. Our results show that FDP improves the rates of low-demand subscribers by 18.4% and charges a penalty of up to 34% to high-demand subscribers for 400 sample observations. This percentage varies with the number of subscribers in the system during an interval and real-time prices.