基于PSO-LS-SVMR的公共建筑能耗短期预测模型

Short-Term Energy Consumption Forecasting Model of Public Building Based on PSO-LS-SVMR

提出一种粒子群算法优化的最小二乘支持向量机回归模型(PSO-LS-SVMR),以实现对公共建筑能耗的短期预测。采用某大型公共建筑物连续31期的用电量及所在地区相关天气指标的实测数据,分别运用PSO-LS-SVMR模型和LMBP神经网络模型对其建筑能耗进行短期预测,并对预测结果展开深入研究。研究结果表明,提出的PSO-LS-SVMR模型在对样本内数据和样本外数据的预测上均取得了较好效果,可以满足公共建筑能耗短期预测的实际需要,为建筑节能管理提供理论支持与决策参考。

To forecast the short-term energy consumption of public building,a novel particle swarm optimization based least squares support vector machine regression model has been proposed. By selecting the monitoring data of a large public building as the original data,i. e. meteorological factors and daily energy consumption,the short-term energy consumption of the large public building was estimated with PSO-LS-SVMR and LMBP neural network separately based on the original data. In order to discover the differences in prediction effect of these models,the predictions were analyzed deeply in accuracy and rationality. The results indicated that the forecasting based on PSO-LS-SVMR is better than others when the samples are known or unknown. Meanwhile,the predictions of short-term public building energy consumption provide the theoretical basis and decision support for the energy-saving management.