摘要
基于山东省1985—2008年的有关数据,分析了山东省电力消费的主要影响因素,然后运用最小-最大规范化方法对数据进行规范化处理,并利用相关系数法得到影响山东省电力需求的最优属性集,最后根据SVM原理,选取径向基核函数、线性核函数、多项式核函数3种核函数分别建立山东省电力需求预测模型,并利用平均绝对百分误差、希尔不等系数和均方根误差3种评价指标进行拟合度的检验。预测结果说明:与传统的时间序列拟合法相比,SVM模型在电力需求预测方面精度较高;在进行短期电力需求预测的时候,基于径向基核函数建立的SVM模型表现出很大的优势;而在进行长期电力需求预测的时候,基于多项式核函数建立的SVM模型更为合适。
Based on the economic data of Shandong province factors influencing electricity consumption, and then uses the from 1985 to 2008, this paper firstly analyzes the main minimum-maximum standardized method to normalize the data, and obtains the optimal attribute set that affects the electric power consumption by using correlation coefficient method. In the end, three kinds of kemel functions （rbf, linear, poly） are used to establish the prediction model based on SVM theory, and the fitting degree is tested with three evaluating indicators （MAPE, PMSE, Theil IC）. The predicted result indicates that compared to the conventional time series fitting method, the SVM model has high accuracy in power demand prediction; the RBF kernel function based SVM model has significant advantages in short-term power demand prediction; while for long-term power demand prediction, the polynomial kernel function based SVM model is more suitable.
出处
《能源技术经济》
2011年第3期40-45,共6页
Electric Power Technologic Economics
基金
国家自然科学基金项目(70871037)
北京市教委共建项目
华北电力大学教改项目
