Load forecasting and operation optimization of residential fresh air system based on artificial neural network

Abstract

Radiant cooling and heating and fresh air system is more and more widely used in residential buildings as a high-comfort, energy-saving and efficient air-conditioning system. The fresh air system handles all the moisture load and part of the cooling load of the building. In actual operation, there are some problems, such as high proportion of energy consumption and mismatch between load and operation characteristics. In this paper, a zone-level artificial neural network (ANN) model is established to predict the moisture load of residential building fresh air system. Compared with the measured data, the zonelevel ANN model is established and verified. The total data used for training and testing are 13260 and 864 respectively. This paper also introduces a system control optimization model, and optimizes the operation of the fresh air system combined with the load forecasting results of the zone-level ANN model. Under the scenario of potential energy storage and time of use price, the optimization control strategy is formulated to improve the flexibility of the system. The results show that the zone-level ANN model has high prediction accuracy. The root mean square error variation coefficients corresponding to the prediction results of moisture load is 8.72%. The optimization results can reduce the operation energy consumption and cost of the system by 27.2% and 29.2% respectively in the whole air conditioning season.