基于自适应深度信念网络的商场空调冷负荷预测

Prediction of Cooling Load of Shopping Mall Air Conditioners Based on Adaptive Deep Belief Network

本文提出了基于自适应深度信念网络的商场空调冷负荷预测方法,通过分析对比本文所提出算法与深度信念网络以及高斯处理后的连续受限玻尔兹曼机-深度信念网络(CRBM-DBN)三种不同的算法模型,得到了可靠的商场空调冷负荷预测模型。通过实验分析,本文改进后的模型在预测准确度方面,较DBN模型均方根相对误差提高66.69%,较CRBM-DBN模型提高55.87%。在运行时间方面,较DBN模型节省了9.11 s,较CRBM-DBN模型节省了6.97 s,模型的收敛速度以及训练精度得到有效提高。

In this paper,a cooling load forecasting method for shopping malls based on adaptive deep belief network is proposed.By analyzing and comparing two different algorithm models of deep belief network and Gaussian processed continuous restricted Boltzmann machines-deep belief network(CRBM-DBN),a reliable forecasting model of air-conditioning cooling load in shopping malls is obtained.Through experimental analysis,in terms of prediction accuracy,the relative root mean square error value of the improved model in this paper is 66.69%higher than that of DBN model and 55.87%higher than that of CRBM-DBN model.In terms of running time,9.11 s is saved compared with DBN model and 6.97 s compared with CRBM-DBN model.Thus,the proposed method can effectively improve the convergence speed and training accuracy of the model.