基于时间序列自回归模型的绿色建筑供暖能耗短期预测

Short-Term Prediction of Heating Energy Consumption in Green Buildings Based on Time Series Autoregressive Model

摘要:

为实时了解绿色建筑供暖能耗的变化趋势,提升能耗预测效果,设计基于时间序列自回归模型的绿色建筑供暖能耗短期预测方法。利用增强迪基-福勒检验法,检验绿色建筑历史供暖能耗时间序列平稳性；对非平稳的历史能耗时间序列进行差分平稳化处理,获取平稳的历史能耗时间序列；在时间序列自回归模型内添加移动平均模型,并考虑能耗的气温影响因素,建立时间序列自回归移动平均模型；利用赤池信息准则确定模型阶数,通过粒子群算法确定模型参数；在模型阶数与参数确定后的模型内,输入平稳的历史能耗时间序列,输出供暖能耗短期预测值。实验证明：该方法可精准预测不同类型绿色建筑的短期供暖能耗；在不同绿色建筑渗透量时,该方法短期供暖能耗预测误差较小；在不同室外温度时,该方法短期供暖能耗预测的可决系数较高,即预测精度较高。

Abstract:

In order to understand the change trend of heating energy consumption in green buildings in real time and improve the prediction effect of energy consumption, a short-term prediction method of heating energy consumption in green buildings based on time series autoregression model is designed. The enhanced Dickie Fowler test is used to test the stability of the historical heating energy consumption time series of green buildings; The non-stationary historical energy consumption time series is processed by difference stationarization to obtain a stable historical energy consumption time series; Add a moving average model to the time series autoregressive model, and consider the temperature influence factors of energy consumption to establish a time series autoregressive moving average model; The order of the model is determined by Akchi information criterion, and the model parameters are determined by particle swarm optimization algorithm; After the model order and parameters are determined, stable historical energy consumption time series are input into the model, and short-term prediction value of heating energy consumption is output. The experiment shows that this method can accurately predict the short-term heating energy consumption of different types of green buildings; The short-term heating energy consumption prediction error of this method is small when the infiltration amount of green buildings is different; At different outdoor temperatures, this method has a high determinable coefficient of short-term heating energy consumption prediction, that is, the prediction accuracy is high.