智能优化LS-SVM在电力负荷预测中的应用

摘要

Short-term power load prediction forecast model has extremely vital significance to the reliability and e-conomical efficiency of power system. Currently, short-term load forecasting algorithm is not enough in prediction ac-curacy , convergence speed and generalization ability. Wavelet packet transform was used to filter signals with noise. Kernel principle conponent analysis and kernel independent component analysis were used in data feature selection. Then chaos particle swarm optimization algorithm was used to determinate optimal parameter in LS-SVM. This intelli-gent optimization algorithm has been tested on short-term power load forecasting and the results show the prediction has higher precision, generalization ability and convergence speed.%研究短期电力负荷的准确预测问题对城市供电系统运行的可靠性和经济性很重要.由于电力负荷变化过程受到各种因素的影响,系统非线性较强.传统方法对电力短期负荷预测精度、收敛速度和泛化能力方面不理想.为解决上述问题,提出了一种核函数KPCA和KICA的最小二乘支持向量机(LS-SVM)电力负荷预测方法.首先对采集的含噪信号进行小波包去噪预处理,同时对不同频段的负荷分量进行区别处理以提高预测精度;利用核函数KPCA和KICA分析方法提取数据特征,提高预测模型的泛化能力和预测精度;采用混沌粒子群算法优化LS-SVM参数提高预测模型的收敛速度.实验表明,改进算法在预测精度、泛化能力及收敛速度方面较突出,具有良好的预测效果.