This paper proposes a novel optimization algorithm, namely hybrid Time Varying Acceleration Coefficients-Gravitational Search Algorithm-Particle Swarm Optimization (hybrid TVAC-GSA-PSO), to solve the large-scale, highly nonlinear, non-convex, non-smooth, non-differential, non-continuous, and complex Combined Heat and Power Economic Dispatch (CHPED) problems. The suggested algorithm is based on the Newtonian laws of gravitation and motion as well as swarm behaviors. The proposed technique combines the best performances of three optimization methods (i.e. GSA, TVAC-GSA, and TVAC-PSO) in terms of their specific strategies in particle movements through a self-adoptive learning strategy. The effectiveness and robustness of the suggested algorithm are tested on a set of five benchmark functions and two large-scale CHPED problems (considering valve-point loading effect and prohibited zones of conventional power units, transmission losses, as well as special characteristic of CHP units). The obtained results by the suggested algorithm in terms of quality solution, computational performance, and convergence characteristic are compared with various algorithms to show the ability of the proposed approach and its robustness in finding a better fuel cost with a less expensive solution. (C) 2017 Elsevier Ltd. All rights reserved.