摘要
针对标准粒子群优化(PSO)算法存在易陷入局部极值点的缺点,提出了一种基于物种概念的动态多种群粒子群优化算法(DMPSO)。在DMPSO中引入了物种概念,在进化过程中动态确定物种,利用种群多样性信息动态调整物种半径,通过物种对解空间的不同区域进行搜索,最终确定出各极值点。将DMPSO算法和支持向量机(SVM)相结合,形成了解决电力系统短期负荷预测问题的新方法(DMPSO-SVM)。在该方法中利用DMPSO算法来优化SVM中的参数,利用快速傅立叶变换(FFT)进行频谱分析并确定SVM的输入量。电力系统短期负荷预测的实际算例表明,与传统预测方法相比,该方法具有更高的预测精度和鲁棒性。
Aiming at the precocious convergence problem of particle swarm optimization algorithm, a dynamic multi- population particle swarm optimization （DMPSO） algorithm is presented. In DMPSO algorithm, the notion of species is used to determine its neighborhood best values and the swarm population is divided into species subpopulations accord- ing to their similarity. Each of these species subpopulations is built around a dominating individual called the species seed. At each iteration step, species seeds are identified from the entire population and then adopted as neighborhood bests for these species groups separately. A strategy for adaptively changing the species radius based on population diversity information is proposed. During iterations, species subpopulations are able to simultaneously optimize toward potentially regions containing multiple optima. A new short-term load forecasting model based on SVM with DMPSO algorithm （DMPSO-SVM） is proposed in which the SVM＇s parameters are optimized by DMPSO algorithm and the input variables of the SVM are determined by fast Fourier transform（FFT）. The example of electricity load data from California power market is used to illustrate the proposed DMPSO-SVM approach. The empirical results reveal that the DMP- SO-SVM approach outperforms the other traditional model.
出处
《计算机科学》
CSCD
北大核心
2008年第7期133-136,172,共5页
Computer Science
基金
国家自然科学基金资助项目(60274009)
教育部博士点基金资助项目(20020145007)
关键词
粒子群优化
动态多种群
物种
支持向量机
负荷预测
Particle swarm optimization, Dynamic multi-population, Species, Support vector machine, Load forecasting
