基于免疫模糊聚类的LSSVM在短期负荷预测中的应用

Application of LSSVM based on immune fuzzy clustering algorithm to short-term load forecasting

最小二乘支持向量机(least square support vector machines,LSSVM)在解决小样本、非线性和高维度问题中表现出许多特有的优势.但是,如果输入的训练数据本身存在着大量的噪声和冗余,LSSVM在训练数据时会因抑制它们而削弱本身的推广能力,结构风险无法达到最小化,从而导致收敛速度慢、预测精度不高等缺点.提出了一种基于免疫模糊聚类(immune fuzzy clustering,IFC)的最小二乘支持向量机预测模型,运用免疫模糊聚类算法对历史数据进行预处理,从聚类后的数据提取LSSVM的训练样本,从而提高训练速度和预测精度,克服LSSVM的上述缺点.最后,将该模型运用到短期电力负荷预测中,与经典的SVM和BP神经网络相比具有更好的泛化性能和预测精度.

Least squares support vector machines has many unique advantages in the performance of solving the small sample, nonlinear and high dimensional problems. However, there are a lot of noise and redundancy in the input training data, and the generalization performance of LSSVM will be weakened due to the noise in the training data, so structural risk minimization can not be achieved, which leads to slow convergence and lower forecast accuracy. This paper presents an immune fuzzy clustering based on LSSVM model, using immune fuzzy clustering algorithm to process pre--history data, extracting the training samples from the clustering data for LSSVM, to improve the training speed and prediction accuracy and overcome the above drawbacks LSSVM. Finally, the model is applied to short-term power load forecasting, and has better generalization performance and prediction accuracy compared to classic SVM and BP neural network.