Incentive-Based Demand Response with Deep Learning and Reinforcement Learning

Incentive-based Demand Response program that can induce end users to reduce power load during peak load period, has been widely implemented due to its flexibility. In this chapter, an incentive-based Demand Response program with modified deep learning and reinforcement learning is presented. A modified deep learning model based on recurrent neural network (MDL-RNN) is first used to identify the future uncertainties of environment by forecasting day-ahead wholesale market price, photovoltaic (PV) power output, and power load. Then, reinforcement learning is utilized to explore the optimal incentive rates at each hour which can maximize the profits of both energy service providers and end users. The results show that the incentive-based DR model contributes to mitigating the supply–demand imbalance and reducing the electricity bills of end users and the expenses of energy service providers. It also shows the potential in implementing incentive-based Demand Response programs under complex and uncertain environment.