电力负荷模型参数辨识的混合优化算法

Hybrid optimization algorithm for load model parameter identification

模拟进化类算法具有全局寻优特性但计算时间过长,而梯度类算法具有很高的局部搜索效率但容易陷入局部最优点。基于模拟进化类算法和梯度类算法的优点提出一种混合优化算法,即以蚁群算法起步,经过一定次数的迭代后切换为梯度算法。提出目标值下降准则和区间收缩准则两种切换算法策略,并且进行对比。针对电力负荷参数辨识,通过仿真算例和实际应用进行测试。结果表明,在保证相同精度的前提下混合优化算法大大提高了计算效率。

The simulated evolutionary algorithm（ SEA） has a global optimization capability,but it requires too much computing time. In contrast,the traditional gradient-based method（ GBM） is characterized by powerful local search efficiency,but it can be easily trapped in a local optima. In order to combine the advantages of both the SEA and GBM,a hybrid optimization algorithm is proposed in this paper. This method starts with the ant colony algorithm（ ACO）; after a certain number of iterations,it switches to the GBM to accelerate the solution procedure.Two switching strategies for the hybrid method,the descent criterion and the interval contraction criterion,are also presented and compared. The hybrid method was applied to the parameter identification of electrical loads. The simulation and practical results show that,in comparison with the ACO,the computational efficiency of the hybrid method can be greatly improved for the same accuracy.