Scholarlink

<p>This study uses regression with Autoregressive Integrated Moving Average (ARIMA) errors method incorporating the COVID-19 lockdown effect to analyse the electricity consumption of a UK college’s facilities. The objective is to accurately predict the electricity consumption of three college campuses (labelled B, C, and T) for use in decision-making activities and planning. As benchmarking methods, Holt-Winters’ multiplicative and additive methods, ARIMA, Neural Network Autoregression (NNAR), and Long Short-Term Memory Recurrent Neural Network (LSTM-RNN) methods are used to predict electricity consumption. The result identified regression with ARIMA errors as the best-performing model, as it considers the effects of the national lockdowns for Coronavirus based on error measures employed. The Mean Absolute Percentage Error (MAPE) values for forecast accuracy are 4.85% for Campus B, 5.31% for Campus C, and 10.76% for Campus T. The proposed method provides a single forecasting equation for all seasons, enables the inclusion of explanatory variables, including an intervention event, and shows good forecasting performance, even though the sample size is not large. The energy used and carbon emission intensities for each campus dropped significantly during the COVID-19 era compared to the pre-pandemic period. This result showed decreases in energy used intensity (EUI) and carbon footprint (CFP) across the three campuses. However, the significant basal energy use highlights a key opportunity for energy savings during low-occupancy periods. The NetZero carbon implication is that this result could be applied to predict carbon footprint of higher education buildings under similar circumstances.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Forecasting energy use and calculating carbon emissions of higher education buildings with COVID-19 lockdown effect

  • 期刊论文

  • 发表日期：

    2026年4月25日

查看全文
  • Blessing, Ibukun Mafimisebi,
  • Bahar Sennaroglu

摘要

This study uses regression with Autoregressive Integrated Moving Average (ARIMA) errors method incorporating the COVID-19 lockdown effect to analyse the electricity consumption of a UK college’s facilities. The objective is to accurately predict the electricity consumption of three college campuses (labelled B, C, and T) for use in decision-making activities and planning. As benchmarking methods, Holt-Winters’ multiplicative and additive methods, ARIMA, Neural Network Autoregression (NNAR), and Long Short-Term Memory Recurrent Neural Network (LSTM-RNN) methods are used to predict electricity consumption. The result identified regression with ARIMA errors as the best-performing model, as it considers the effects of the national lockdowns for Coronavirus based on error measures employed. The Mean Absolute Percentage Error (MAPE) values for forecast accuracy are 4.85% for Campus B, 5.31% for Campus C, and 10.76% for Campus T. The proposed method provides a single forecasting equation for all seasons, enables the inclusion of explanatory variables, including an intervention event, and shows good forecasting performance, even though the sample size is not large. The energy used and carbon emission intensities for each campus dropped significantly during the COVID-19 era compared to the pre-pandemic period. This result showed decreases in energy used intensity (EUI) and carbon footprint (CFP) across the three campuses. However, the significant basal energy use highlights a key opportunity for energy savings during low-occupancy periods. The NetZero carbon implication is that this result could be applied to predict carbon footprint of higher education buildings under similar circumstances.