Optimizing Dynamic Pricing Demand Response Algorithm Using Reinforcement Learning in Smart Grid

The emergence of the smart grid has presented new opportunities for demand response to control and reduction of electricity usage. With sufficient energy management, in off-peak times customers can buy and store more electrical energy than on-peak times or shift the shiftable appliance to proper times. These scenarios provide the desired change in the load profile shaping which is the main function of demand response. Dynamic pricing is one of the most effective ways to encourage customers to change their consumption pattern. However, it is difficult to determine an optimal pricing strategy because of the uncertainty of electricity consumption. In this paper, we used NeuralBandit algorithm to solve this problem, which using a neural network that learns the context and the associated reward. Simulation result shows that this proposed algorithm can reduce energy cost, improve reliability of the system and control the ramp rate of the power system.