自适应加权最小二乘支持向量机的空调负荷预测

发布时间：2018-10-20 10:45

【摘要】：为了提高建筑空调负荷的预测精度,在分析空调负荷主要影响因素的基础上提出了一种基于自适应加权最小二乘支持向量机(AWLS-SVM)的建筑空调负荷预测方法。该方法根据预测误差的统计特性,采用基于改进正态分布加权规则,自适应地赋予每个建模样本不同的权值,以克服异常样本点对模型性能的影响。建模过程中采用粒子群优化(PSO)算法对模型参数进行优化,以进一步提高模型预测精度。基于DeST模拟数据将AWLS-SVM方法应用于南方地区某办公建筑的逐时空调负荷预测中,并与径向基神经网络(RBFNN)模型、LS-SVM模型及WLS-SVM模型作比较,其平均预测绝对误差分别降低了51.84%、13.95%和3.24%,并进一步基于实际空调负荷数据将该方法应用于另一办公建筑的逐日空调负荷预测中。预测结果表明:AWLS-SVM预测的累积负荷误差为4.56MW,亦优于其他3类模型,证明了AWLS-SVM具有较高的预测精度和较好的泛化能力,是建筑空调负荷预测的一种有效方法。
[Abstract]:In order to improve the precision of building air conditioning load prediction, an adaptive weighted least square support vector machine (AWLS-SVM) based building air conditioning load forecasting method is proposed based on the analysis of the main factors affecting the air conditioning load. According to the statistical characteristics of prediction error, the method adaptively assigns different weights to each modeling sample based on the improved normal distribution weighting rule, in order to overcome the influence of abnormal sample points on the performance of the model. In the process of modeling, particle swarm optimization (PSO) algorithm is used to optimize the model parameters to further improve the prediction accuracy of the model. Based on the DeST simulation data, the AWLS-SVM method is applied to the hourly air conditioning load forecasting of an office building in southern China. The results are compared with the radial basis function neural network (RBF) (RBFNN) model, LS-SVM model and WLS-SVM model. The average absolute error of prediction is reduced by 51.84% and 3.24%, respectively. The method is further applied to the daily air conditioning load forecasting of another office building based on the actual air conditioning load data. The prediction results show that the cumulative load error of AWLS-SVM is 4.56MW, which is better than other three models. It is proved that AWLS-SVM has higher prediction accuracy and better generalization ability, and it is an effective method for building air conditioning load forecasting.
【作者单位】：福州大学石油化工学院;
【基金】：国家自然科学基金资助项目(6080402,61374133) 高校博士点专项科研基金(20133314120004)~~
【分类号】：TU831.1;TP18

【相似文献】