Cuckoo search(CS) is a novel nature-inspired algorithm. For the sake of adaptation to various optimization problems, an improved CS algorithm is proposed, named swarm feature feedback cuckoo search(SFFCS) algorithm, on the basis of the feedback control principle. Swarm features such as age structure and success rate of mutation are introduced as feedback information for adjusting the parameters dynamically. Double evolutionary strategies and strategy selection probability are also introduced to balance the capability between local and global search. Numerical experiments on benchmark functions and the optimal power flow problem of the electrical system are conducted. The results indicate that the SFFCS algorithm behaves strong performance on convergence and adaptation, and show the effectiveness and practical engineering value of the proposed algorithm.