摘要
考虑到实际电力负荷预测中各数据的重要程度并不相同,在标准最小二乘支持向量机回归算法的训练样本中设置权值系数,建立了加权最小二乘支持向量机模型,以实现样本的优化选择,达到历史数据"重近轻远"的学习效果;同时考虑到粒子群优化算法收敛速度快和混沌运动遍历性、随机性等特点,提出了一种基于混沌思想的粒子群优化算法对模型参数进行优化,引入优势粒子和劣势粒子的权重自适应调节机制,使算法具有动态适应性。将改进的模型应用于江西省萍乡市月度负荷预测中,结果表明本文方法与常规方法相比降低了预测误差,且速度较快。
This paper proposes an improved least square support vector machine for monthly load forecasting.On the one hand,the model can implement optimization selection,determine the weight proportions with history data;On the other hand,based on the ergodicity and randomness of chaotic motion,an improved chaotic PSO algorithm is proposed to optimize model parameters.Then,the weight self-adaptive adjustment mechanism for dominant particles and bad particles is introduced and it can make algorithm easily jump out of local optimum with effective dynamic adaptability.Finally,improved model is applied to forecast monthly load in Pingxiang City of Jiangxi Province.Compared with the traditional methods,the results show that the proposed method can reduce forecasting error and it has better performance.
作者
吴钰
王杰
WU Yu;WANG Jie(Department of Electrical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China)
出处
《水电能源科学》
北大核心
2012年第5期174-177,共4页
Water Resources and Power
基金
国家自然科学基金资助项目(61074042)
关键词
月度负荷
预测
最小二乘支持向量机
加权
混沌粒子群
monthly load
forecasting
least square support vector machine
weighted
chaotic particle swarm optimization
