神經網絡在不同供暖模式下建筑能耗預測中的應用

摘 要：【目的】為進一步優化供暖系統的運行模式，降低系統能耗，提高能源利用率。【方法】利用室外溫度、日期類型、歷史能耗等數據進行動態建模，分別建立了基于前饋（Back Propagation，BP）神經網絡和長短期記憶（Long Short-Term Memory，LSTM）神經網絡的能耗預測模型，對連續運行和間歇運行模式下的供暖系統的能耗進行預測。【結果】將預測值與真實值進行對比，結果表明，在連續運行模式下，供暖系統的能耗預測精度總體高于間歇運行模式。在間歇供暖運行模式下，LSTM神經網絡的預測結果更好。在連續供暖的運行模式下，BP神經網絡能獲得更好的預測結果。【結論】不同預測模型適合不同供暖模式，在對能耗進行預測時，需要選擇相匹配的預測模型，能提高能耗預測的精度。

關鍵詞：能耗預測；前饋神經網絡；長短期記憶神經網絡

中圖分類號：TK01" " " 文獻標志碼：A" " 文章編號：1003-5168（2023）21-0023-04

DOI：10.19968/j.cnki.hnkj.1003-5168.2023.21.005

Application of" Neural Network in Building Energy Consumption

Prediction under Different Heating Modes

WU Yunhe

（ Zhengzhou Branch of China Railway Fifth" Survey and Design Institute Group Co.， Ltd.， Zhengzhou 450000，China）

Abstract： [Purposes] This paper aims to further optimize the operation mode of the heating system， reduce the energy consumption of the system and improve the energy utilization rate. [Methods] The outdoor temperature， date type， historical energy consumption and other data were used for dynamic modeling. The energy consumption prediction models based on Back Propagation （ BP ） neural network and Long Short-Term Memory （ LSTM ） neural network were established respectively to predict the energy consumption of heating systems under continuous operation and intermittent operation modes. [Findings] The predicted values were compared with the true values. The results showed that the energy consumption prediction accuracy of the heating system in the continuous operation mode was generally higher than that in the intermittent operation mode. In the intermittent heating operation mode， the prediction results of LSTM neural network are better. In the operation mode of continuous heating， BP neural network can obtain better prediction results. [Conclusions] Different prediction models are suitable for different heating modes. When predicting energy consumption， it is necessary to select a matching prediction model to improve the accuracy of energy consumption prediction.

Keywords： energy consumption prediction; feedforward neural network; long short-term memory neural network

0 引言

Kawashima等^［1］利用人工神經網絡（Artificial Neural Network，ANN）預測出中央空調系統的能耗，并將預測結果應用于空調系統的運行控制中，結果表明，由ANN預測控制的系統能耗減少6.9%、運行費用降低13.5%。Deb等^［2］利用ANN神經網絡建立了能耗預測模型，并預測了某辦公建筑的空調能耗，平均絕對誤差為14.8%。Nivethitha等^［3］利用卷積神經網絡（Convolutional Neural Network，CNN）對能耗產生影響的非線性交互特征進行分析，并建立基于KCNN-LSTM的深度學習模型。Safa等^［4］建立了基于多元線性回歸及ANN模型，預測出新西蘭地區某建筑能耗，通過ANN模型預測來降低該建筑能耗，并提高系統運行的穩定性。Qing等^［5］利用天氣預報數據對日前照度進行預報，用LSTM算法預測時的均方根誤差比BP神經網絡降低42.9%。……