This paper summarizes the community where people live development laws and proposes a novel meta-heuristic called community development algorithm (CDA). To enhance the optimization process, CDA innovatively designed two unique attraction effects based on the degree of people’s direct contribution to society and abstracted four novel inter-particle learning ways based on people’s four social activities: Marriage, Entrepreneurship, Working, and University. Theoretically, it is demonstrated that CDA can satisfactorily balance the exploration and the exploitation stages, and necessary and sufficient conditions for CDA convergence are derived. The 59 functions of Congress on Evolutionary Computation 2014 and 2017 are introduced as the benchmark. Under the pressure of other 32 well-known meta-heuristics, statistical results, including averages, standard deviations, rank-sum tests, and sign-rank tests, as well as graphical results, including convergence curves and box plots, confirm that CDA is competent for solving numerical optimization problems with efficiency and reliability. Significantly, the quantitative analysis demonstrated that CDA’s average ranking is ahead of the emperor penguin optimizer by 73.5% while ahead of the fully informed search algorithm by 62.1%, respectively. Additionally, the scalability and fastness of CDA are validated by scalability and wall clock time analysis, and the classical application of Congress on Evolutionary Computation 2011 demonstrates the outstanding accuracy and stability of CDA. Finally, CDA is utilized to optimize the controller parameter in the main steam temperature control system of the supercritical unit. The control response curves show that CDA generally provides better parameters than compared algorithms, demonstrating the merits and potential to solve complex real-world engineering problems.
Read full abstract