Time of Use Tariff parameter Estimation: A Data Analysis Approach on Multicore Systems

Abstract

Increased use of solar energy is forcing energy companies to devise new time of use (ToU) tariff scheme to counter revenue losses. Designing ToU tariff scheme is a complex multi-stage problem. The adoption of smart meters and availability of high performance multicore systems has opened up newer and better ways of tariff design. The design of ToU tariff schemes typically involves identifying various demand periods which is accomplished by analyzing the intraday consumption patterns across various geographies. The optimal tariff parameters for all the demand periods is then computed by solving a constrained optimization problem. In this paper, we present a present multi dimensional grid search approach to compute the optimal tariff parameters for a ToU scheme. The grid search method was then efficiently implemented on Nvidia GPUs with dynamic parallelism. The annual energy consumption data processing for nearly 0.3 million consumers and computation of consumption pattern and demand periods was carried out using MPI based parallel processing on an Intel Haswell system.