由于被动式近零能耗建筑实际日耗热量受多种因素影响、特征难提取，导致日耗热量预测难度较大。为此，提出一种基于离散稀疏函数的建筑实际日耗热量预测方法。采用离散稀疏函数计算历史建筑日耗热量数据，在不同维度层次上特征向量和稀疏参数，利用激活函数建立偏离惩罚项，明确每个热量先验信息数据与中心值间的偏离度。采用线性传递函数求得会影响实际日耗热量间的线性变化关系，建立时间序列，采用自回归算法得出时间和热量的正向变化序列，实现对日耗热量的预测。实验数据证明，所提方法日耗热量预测精准度较高，针对热负荷、冷负荷以及预测平均评价(Predicted Mean Vote，PMV)指标均实现了高效预测。
Simulation of Daily Heat Consumption Prediction for Passive Near-Zero Energy Building
Actually,the daily heat consumption of passive near-zero energy buildings is influenced by various factors,making feature extraction difficult,so it is difficult to predict the daily heat consumption.To address this,a method for predicting the actual daily heat consumption of buildings based on discrete sparse functions was proposed.Firstly,discrete sparse functions were employed to calculate historical daily heat consumption data,thus obtaining the feature vectors and sparse parameters on different dimensions.Moreover,activation functions were used to establish a deviation penalty term,thus determining the deviation between each prior heat information data and the central value.Furthermore,a linear transfer function was used to determine the linear change relationship affecting the actual daily heat consumption,and then a time series was constructed.Finally,an autoregressive algorithm was adopted to derive the positive sequence of time and heat,thereby achieving the prediction of daily heat consumption.Experimental data proves that the proposed method has a high accuracy in predicting daily heat consumption.It has achieved efficient predictions for thermal load,cooling load,and Predicted Mean Vote(PMV)indicators.
Passive near zero energy buildingActual daily heat consumptionDeviationDiscrete sparse func-tionHeat penetration by hot air
NETL
NSTL
万方数据
