由于建筑能耗模型多峰值的特点，使用粒子群算法处理建筑节能优化问题时，容易在局部最优区域过早收敛，将分布式的变异算子与粒子群算法结合可对此进行改进。文中研究了 4 种现有的粒子群变异算法，针对其优化效果的不足，对发生变异的具体操作进行了改进，使用测试函数验证了其有效性，并将其应用到建筑节能优化问题当中。经过大量实验发现，对于文中的建筑节能优化问题，改进的粒子群变异算法相对于现有的粒子群变异算法，其目标函数 (即建筑的太阳能辐射得热量) 的平均值下降了1. 3%以上，收敛率至少提高了 3 倍，寻优效果有明显的改善，证明所提出的改进算法具有有效性和普适性，可以在一般的建筑节能优化问题中推广。

When particle swarm optimization algorithm is used in the building energy-saving optimization，it is easy to converge untimely in the local optima due to the multi-peak characteristic of the building energy model． The combination of the distributed mutation operator and particle swarm algorithm can provide a solution to this pro-blem． In this paper，4 existing particle swarm mutation algorithms were studied． Aiming at the problem of low effects of optimization，the specific operations of the mutations were improved，and the effectiveness of 4 improved algo-rithms was verified by test functions． Then，they were applied to a building energy-saving optimization case． Experi-mental results show that，as compared with the existing particle swarm mutation algorithms，the improved algorithm can decrease the average value of object function by 1. 3% at least and increase the convergence rate by over 3 times，which means that the optimization effect is obviously improved． Thus，it is concluded that the improved al-gorithms is effective and applicable in general building energy-saving optimization．