Multi-Step Short-Term Building Energy Consumption Forecasting Based on Singular Spectrum Analysis and Hybrid Neural Network

Short-term building energy consumption forecasting is vital for energy conservation and emission reduction. However, it is challenging to achieve accurate short-term forecasting of building energy consumption due to its nonlinear and non-stationary characteristics. This paper proposes a novel hybrid short-term building energy consumption forecasting model, SSA-CNNBiGRU, which is the integration of SSA (singular spectrum analysis), a CNN (convolutional neural network), and a BiGRU (bidirectional gated recurrent unit) neural network. In the proposed SSA-CNNBiGRU model, SSA is used to decompose trend and periodic components from the original building energy consumption data to reconstruct subsequences, the CNN is used to extract deep characteristic information from each subsequence, and the BiGRU network is used to model the dynamic features extracted by the CNN for time series forecasting. The subsequence forecasting results are superimposed to obtain the predicted building energy consumption results. Real-world electricity and natural gas consumption datasets of office buildings in the UK were studied, and the multi-step ahead forecasting was carried out under three different scenarios. The simulation results indicate that the proposed model can improve building energy consumption forecasting accuracy and stability.

Automated summary

In this study, a novel hybrid model SSA-CNNBiGRU is proposed for short-term building energy consumption forecasting. The model integrates singular spectrum analysis, convolutional neural network, and bidirectional gated recurrent unit neural network. By decomposing, feature extraction, and time series forecasting of the energy consumption data, the proposed model achieves more accurate and stable energy consumption prediction.