Reinforcement learning-enhanced multi-strategy cuckoo search algorithm to solve combined economic emission dispatch problems

The combined economic emission dispatch (CEED) problem of balancing harmful gas emissions and power generation costs is of paramount importance. In this regard, this study proposes a reinforcement learning (RL) enhanced multi-strategy cuckoo search (RLMCS) algorithm for solving the CEED problem. Through a price penalty factor, this paper transforms the economic and emission objectives into a single-objective optimization and introduces three search strategies into the traditional cuckoo search (CS): multidimensional learning (ML), one-dimensional learning (OL), and stochastic learning (SL). ML helps to circumvent the local optimum trap; OL improves the algorithm’s search efficiency in the optimal direction; and SL enhances the algorithm’s global search capability. In addition, the RL technique is used to dynamically regulate the multi-policy mechanism, guiding the optimal policy selection and ensuring that the algorithm uses the most efficient updating method at different stages. The algorithm is experimentally validated on 29 benchmark tests and CEED problems with 6 and 11 units under different load demands. The RLMCS shows a high degree of robustness when solving CEED problems.