基于预测的复合地源热泵系统控制方法研究

Research on the Control Method of the Hybrid Ground Source Heat Pump System Based on Prediction

在复合式地源热泵系统中控制策略存在着极大的优化空间,提出一种新的更为有效的控制方法,即在并联系统中直接比较冷却塔和地埋管出口温度的方法,然而在实际运行中只能实时测得一个出口水温,因此需要建立一个可靠的土壤换热器模型预测其出口水温。运用人工神经网络(ANN)实现该做法,利用FLUENT软件模拟动态复合式地源热泵系统为ANN模型提供训练、测试样本。为获得最优模型,土壤换热器的人工神经网络进行优化。结果表明,在LM算法下,隐层神经元数目为14的网络结构最为理想,预测结果绝对误差不超过0.15℃。

A new control strategy in hybrid ground source heat pump systems（HGSHP） coupled with cooling tower is proposed,that is,to compare water temperatures exiting the cooling tower and ground heat exchanger（GHE） directly.However,only one of the two temperatures can be measured timely so it is necessary to build a predictive model.Artificial neural network（ANN） is introduced to make it.A dynamic numerical simulation of hybrid ground source heat pump system based on FLUENT is employed to generate the monitoring data which are then used to train and test ANN models.Furthermore,the optimal model is obtained through the optimization of several parameters.Results show that ANN models can be used to predict the water temperature at outlet of the GHE exactly and the model with 14 neurons in hidden layer,trained with LM algorithms is optimal.