基于支持向量回归机的空调逐时负荷滚动预测算法

来源期刊:中南大学学报(自然科学版)2014年第3期

论文作者:周璇 杨建成

文章页码:952 - 958

关键词:空调逐时负荷;滚动预测算法;支持向量回归机;网格搜索遍历算法;期望误差

Key words:hourly load of air conditioning system; rolling forecasting algorithm; support vector regression; mesh optimization algorithm; expected error

摘    要:针对当前空调负荷预测算法精度不高难以满足空调系统节能优化控制的问题,提出基于支持向量回归机(Support Vector Regression,SVR)的空调逐时负荷滚动预测算法,建立SVR滚动预测模型,模型参数采用网格搜索遍历算法进行寻优。为简化模型的复杂性,还对影响空调负荷的主要因素进行了相关性分析。此外,算法利用当日前1 h的滚动信息,不断对模型进行修正以提高负荷预测精度。最后探讨以期望误差为预测精度评价指标时,不同训练样本长度对神经网络和SVR算法预测精度的影响。预测结果表明:基于支持向量回归机的空调逐时负荷滚动预测算法较BP神经网络算法的预测精度提高10.3%,比常规支持向量回归机算法预测精度提高23.9%,训练样本较小时,算法预测性能更为优越。

Abstract: For load prediction algorithm accuracy of air conditioning system is not high enough to fulfill the requirement of energy optimal control for energy conservation, a support vector regression (SVR) model and its algorithm for hourly load rolling forecasting were proposed in this paper, with the parameters optimized by mesh optimization algorithm. Moreover, the correlation analysis was used to analyze the main factors affecting the load of air conditioning system to simplify the model and the previous 1-hour load and weather information were used to modify the model in the forecasting process continuously. Furthermore, the effect of training-sample size on prediction accuracy was discussed. Finally, as expected error percentage was adopted as the evaluation of the prediction accuracy, the results showed that the prediction accuracy of rolling forecasting algorithm based on support vector regression was 10.3% and 23.9% more accurate than BP neural network algorithm and conventional SVR algorithm respectively when the number of the training set is 3 672, which was better with little training sample set.

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号