WCSE

Forecast of Building Energy Consumption Using RBF Neural Network based on L-GEM

Abstract— Building energy consumption accounts for a large part of the total energy consumption of the society. Its accurate prediction is very important for energy-saving. In order to forecast building energy consumption, RBF neural network (RBFNN) is used. Based on the discussion that the neural networks yield small error on training dataset does not always perform well on testing dataset, we proposed a method to use RBFNN by applying the Localized Generalization Error Model (L-GEM) in this paper. To obtain the optimum localized generalization error (L-GE), a mofified chaotic mutation PSO is presented to search the optimum L-GE of RBFNN. Experiments have been carried and the results show that the RBFNN based on the L-GEM outperforms general RBFNN in building energy consumption prediction.

Index Terms— building energy consumption, RBF neural network; localized generalization error model.

Yuncan Xue, Enpeng Qiu, Juntao Fei
College of IoT Engineering, Hohai University, CHINA

[Download]

Cite: Yuncan Xue, Enpeng Qiu, Juntao Fei, "Forecast of Building Energy Consumption Using RBF Neural Network based on L-GEM," Proceedings of 2018 the 8th International Workshop on Computer Science and Engineering, pp. 351-355, Bangkok, 28-30 June, 2018.