摘要
针对现有短期电力负荷预测局部优化模型预测准确度不高的问题,提出了一种全过程优化的支持向量机(Support Vector Machine,SVM)模型用于短期电力负荷预测。该模型采用全过程优的建模思想从3个方面对支持向量机进行优化,首先采用模糊C均值聚类算法对输入特征集进行处理;然后采用了组合核函数作为支持向量机的核函数;最后由于烟花算法具有爆发性、隐并行性、多样性和瞬时性等优点,将其用于SVM的核函数参数与惩罚系数C的优化选取中。预测结果表明没有经过优化的支持向量机平均绝对百分误差(Mean Absolute Percentage Error,MAPE)为4. 17%,仅采用烟花算法进行局部优化的平均绝对百分误差为3. 25%,而全过程优化支持向量机的为2. 28%。提出的全过程优化支持向量机模型有效地提高了短期电力负荷预测预测的准确度。
To improve the prediction accuracy of the existing local optimization short-term load forecasting model, a whole process optimization support vector machine（GP-SVR） model is proposed in this paper. The model uses the whole process optimization modelling idea to optimize the support vector machine from three aspects. Firstly, this model uses the fuzzy C -means clustering algorithm to dealing with the input character sets. Then, the compound kernel function is chosen in this paper. Finally, because the fireworks algorithm has the advantages of explosiveness, implicit parallelism, diversity, and instantaneity, this model applied it to optimize the penalty coefficient C and the parameters of the kernel function. The prediction results show that the mean absolute percentage error（MAPE） of the SVM without optimization is 4.17% and the MAPE of the SVM based on fireworks algorithm is 3.25%. As for the whole process optimization SVM, the MAPE is 2.28%. The whole process optimization support vector machine model proposed in this paper effectively improves the accuracy of short-term power load forecasting.
作者
简献忠
顾祎婷
JIAN Xianzhong;GU Yiting(Ministry of Education and Shanghai,Key Lab of Modern Optical System,Electrical Engineering College,University of Shanghai for Science and Technology,Shanghai 200093,China)
出处
《电力科学与工程》
2018年第11期45-51,共7页
Electric Power Science and Engineering
基金
国家自然科学基金资助项目(41075019)
关键词
全过程优化
支持向量机
电力负荷预测
烟花算法
whole process optimization
support vector machine
electric load forecasting
fireworks algorithm
