DED (Dynamic economic dispatch) considering valve-point effects is a complicated non-convex optimization problem in power system. For large-scale DED problem with hundreds of generators, how to avoid the curse of dimensionality remains a big challenge due to the exponential growth of search space. In allusion to this problem, this paper presents a promising heuristic approach named CSO (crisscross optimization) algorithm, which generates high quality solutions in large space by applying two interacting search operators, namely horizontal crossover and vertical crossover. The former has powerful global search ability while the latter can effectively alleviate the premature convergence problem. Their combination leads to a magical effect on improving solution quality and convergence rate especially for large-scale DED problems with valve-point effects. The feasibility and effectiveness of the proposed CSO algorithm is validated by seven test systems consisting of different numbers of generators. The results are compared with those of other heuristic methods reported in the literature. It is shown that the proposed method is capable of yielding higher quality solutions. To examine the availability of CSO in large power system, three new systems with 200-1000 generators are also tested, the obtained results confirm its suitability for large-scale DED problem. (C) 2015 Elsevier Ltd. All rights reserved.