A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Short-Term Electric Load Forecasting with a Hybrid ARIMA, SVR, and IA Methodology
Accurate prediction of short-term electric load plays an indispensable role in improving real-time control and operation of building energy. However, due to various influencing factors such as building thermal environment, weather conditions, and operating hours, there is huge degree of uncertainty and randomness in historical building electric load data containing both linear and nonlinear components, which makes predicting short-term electric load accurately a challenging work. To improve the accuracy of building electric load prediction, the paper proposes a hybrid model incorporating autoregressive integrated moving average (ARIMA) with support vector machine (SVR) optimized by the immune algorithm (IA), aiming to grasp the respective merits of the ARIMA and the SVR. Among them, ARIMA is used to predict linear component of electric load series while SVR is used to predict the nonlinear component of the electric load series. Then, by integrating the prediction results of the two models, the final predicted value of the short-term electric load series is obtained. To examine the precision of the proposed model, experiment on the electricity load (half-hourly data) of a general hospital building in Shanghai obtained from the intelligent building energy support system (i-BESS) of Shanghai Hospital Development Center (SHDC) is conducted. The performance of the hybrid load prediction is verified using statistical metrics, namely, mean absolute percentage error (MAPE) and the root mean square error (RMSE). Empirical results indicate that ARIMA-SVR2 model outperforms other models, including single models (ARIMA and SVR) and the other two hybrid models (ARIMA-SVR1 and ARIMA-SVR3), displaying lowest MAPE and RMSE (0.85% and 4.47). Therefore, the proposed hybrid model can be promoted as a useful tool for efficient prediction of short-term electric load in the future.
Short-Term Electric Load Forecasting with a Hybrid ARIMA, SVR, and IA Methodology
Accurate prediction of short-term electric load plays an indispensable role in improving real-time control and operation of building energy. However, due to various influencing factors such as building thermal environment, weather conditions, and operating hours, there is huge degree of uncertainty and randomness in historical building electric load data containing both linear and nonlinear components, which makes predicting short-term electric load accurately a challenging work. To improve the accuracy of building electric load prediction, the paper proposes a hybrid model incorporating autoregressive integrated moving average (ARIMA) with support vector machine (SVR) optimized by the immune algorithm (IA), aiming to grasp the respective merits of the ARIMA and the SVR. Among them, ARIMA is used to predict linear component of electric load series while SVR is used to predict the nonlinear component of the electric load series. Then, by integrating the prediction results of the two models, the final predicted value of the short-term electric load series is obtained. To examine the precision of the proposed model, experiment on the electricity load (half-hourly data) of a general hospital building in Shanghai obtained from the intelligent building energy support system (i-BESS) of Shanghai Hospital Development Center (SHDC) is conducted. The performance of the hybrid load prediction is verified using statistical metrics, namely, mean absolute percentage error (MAPE) and the root mean square error (RMSE). Empirical results indicate that ARIMA-SVR2 model outperforms other models, including single models (ARIMA and SVR) and the other two hybrid models (ARIMA-SVR1 and ARIMA-SVR3), displaying lowest MAPE and RMSE (0.85% and 4.47). Therefore, the proposed hybrid model can be promoted as a useful tool for efficient prediction of short-term electric load in the future.
Short-Term Electric Load Forecasting with a Hybrid ARIMA, SVR, and IA Methodology
Li, Yongkui (author) / Cao, Lingyan (author) / Han, Yilong (author) / Shi, Yuchen (author) / Zhang, Yan (author)
Construction Research Congress 2020 ; 2020 ; Tempe, Arizona
Construction Research Congress 2020 ; 166-175
2020-11-09
Conference paper
Electronic Resource
English
Short-term load forecasting coupled with weather profile generation methodology
| SAGE Publications | 2018
Hybrid Models Combining EMD/EEMD and ARIMA for Long-Term Streamflow Forecasting
| DOAJ | 2018
Short-term electric load forecasting using fuzzy neural networks
| British Library Online Contents | 1995