Optimal scheduling of a cogeneration system via Q-learning-based memetic algorithm considering demand-side response

Abstract

An optimization scheduling method for cogeneration systems based on the Q-learning-based memetic algorithm (QMA) is developed to enhance the system's wind power absorption capacity and economical operation. First, an optimal scheduling model for cogeneration systems is constructed. The model integrates time-of-use electricity pricing and heating comfort into the demand side, fully exploiting the potential of demand-side response (DR) in power generation and heating. An additional heat source (AHS), which consists of an electric boiler and heat storage tank, is used to uncouple heat and power. In addition, a QMA algorithm is designed to find the optimal operating scheme for the cogeneration system. The Q-learning algorithm is introduced to dynamically adjust crossover and mutation parameters during the global evolution stage, improving the algorithm's search capability. The Taguchi method is utilized for algorithm parameter calibration. Finally, the simulation results under various operating scenarios are compared and analyzed, verifying the feasibility and effectiveness of the proposed method. The simulation results show that DR and AHS can improve the system's economic performance and wind power utilization rate. Compared with the IMA, IPSO, MA, and IABC algorithms, the QMA algorithm reduces the average economic cost by approximately 1.18 %, 2.23 %, 5.69 %, and 14.69 %, respectively.