Generally, differential evolution (DE) algorithm is easily stuck into local optima as well as suffers from low convergence accuracy when employed for optimization of heat exchanger network. To solve these issues, an opposition-based multi-population parallel differential evolution algorithm is proposed. Firstly, opposite population is built by using initial population. Then, new generation of individuals are generated through information exchange, which is produced by mutated operation between opposite population and its original correspondence. The final step is to retain evolution of multi-population in parallel by applying multi-round opposites, so that the population is enable to keep current solution information and search new solutions in a larger space as well. Computing results of improved DE algorithm on 9sp and 15sp suggests that the method improves population diversity, jumps out local optima and at the same time achieves higher speed and accuracy.

For optimization problem of heat integration system, serious nonlinear and non-convex heat exchanger network belonging to multi-extremum and multi-dimensional problems are considered. Dynamic multi-agent differential evolution algorithm is provided to solve the problem. It makes use of sensing capability of multi-agent with dynamic update strategy, which improves formation mechanism of population and mutation mechanism of differential evolution algorithm and globle searching ability in large scale nonlinear system. The algorithm was applied to 10SP2 and 9SP1 cases of heat exchanger network problems. Better total annual cost is obtained, which indicates better globle searching ability of the algorithm.