Optimization of the refrigerant circuit (RC) of a fin-and-tube heat exchanger can increase its heat exchange capacity or decrease its cost. The genetic algorithm is one of the suitable optimization methods, however it needs to be improved for RC optimization of fin-and-tube heat exchangers. An improved genetic algorithm (IGA) is proposed for RC optimization. In the IGA, the RC solutions are represented by one-dimensional integer strings which can save both computer memory and decoding time. RC correction operators are developed and embedded in the entire genetic process with the goal of avoiding physically impossible solutions. The knowledge-based RC generation method, greedy RC crossover method, greedy RC mutation method and all-previous-population based selection method are developed in order to improve the efficiency of the genetic evolution process for RC optimization. Case studies with 3 different heat exchangers show that both the optimization speed and the quality of the output optimal solution of IGA are better than those of the conventional genetic algorithm. A 0–40% decrease in total length of joint tubes is obtained after optimization with the IGA with the target of obtaining the shortest joint tubes. In addition, a 2.8–7.4% increase in heat exchange capacity is obtained after IGA optimization with the target of maximum heat transfer.