技术中心

摘 要： 空调系统负荷非线性变化，单独利用人工神经网络方法进行预测，受到样本数量少等的限制，精度一般都比较低。空调系统负荷时间序列包含了参与空调系统动态变化的全部变量的信息，可以利用相空间重构技术提取和恢复出系统原来的规律。重构后的相空间具有与实际动力系统相同的几何性质，这样就可以大大增加样本数。相空间重构受嵌入维数m和延迟时间的影响，确定这两个重构参数的最佳数值是非常重要的。建立了基于相空间重构技术的神经网络空调负荷预测模型，在嵌入维数m和延迟时间分别取不同的值时，利用这个模型对相同时间段的空调负荷进行预测，选择预测误差最小时对应的参数为最佳相空间重构参数，对应的预测值为最佳预测结果。结果表明，这个方法有很好的效果，具有简化计算复杂性等特点。 关键词 ：空调系统；负荷；相空间重构；人工神经网络；预测 Load Predict For Hvac System Based On Optimum Parameters Of Phase Space Reconstruction Zhu Yonghua Du Zhimin Yang Xuebin He Keqing Chen Ping Jin Xinqiao School of mechanical engineering, Shanghai Jiao Tong University, Shanghai 200240, China Abstract: Artificial neural network can be used to predict nonlinear changes in air-conditioning load. The precision, however, is poor due to fewer samples. For air-conditioning system, the load time series contains all variables information reflecting the dynamic changes of air-conditioning systems. The phase space reconstruction system is used to extract and recover the original law. Compared with the actual dynamic system, the reconstructed phase space has the same geometric properties, so the number of samples is greatly increased. Determining two parameters, embedding dimension m and delay time , is very important for phase space reconstruction. The neural network predicting model is developed on the basis of phase space reconstruction for air conditioning load. The air conditioning load is predicted under different embedding dimension m and delay time . The results show that the method can simplify the computational complexity. Keywords : air conditioning system; load; phase space reconstruction; artificial neural network; predict