A combination approach based on seasonal adjustment method and echo state network for energy consumption forecasting in
- PDF / 2,880,616 Bytes
- 20 Pages / 547.087 x 737.008 pts Page_size
- 33 Downloads / 135 Views
ORIGINAL ARTICLE
A combination approach based on seasonal adjustment method and echo state network for energy consumption forecasting in USA Lan Qin & Weide Li
Received: 14 July 2019 / Accepted: 26 August 2020 / Published online: 4 September 2020 # Springer Nature B.V. 2020
Abstract It is extremely significant to construct a scientific and accurate forecasting model for energy consumption of the USA, because it could help to formulate energy policies and allocate energy resources. In recent years, more and more hybrid models based on dividedand-conquer method have been applied in energy consumption prediction to obtain satisfactory results. However, owning to the obvious enhancing effect of decomposition methods, the issue concerning seasonal fluctuation existed in time series is rarely considered before modeling. It is a fact that seasonality indeed influences the performance of prediction. This paper proposes a hybrid forecasting model for energy consumption, which combines seasonal adjustment method, ensemble empirical mode decomposition (EEMD), echo state network (ESN), and grasshopper optimization algorithm (GOA). The seasonal adjustment method that inherits the idea of divide and conquer is used to decompose original time series into only two parts including seasonal subseries and remainder subseries rather than regular three parts(seasonality, trend, and residual) for avoiding the complex modeling task of the residual subseries. Then models ESN and EEMD-GOA-ESN
L. Qin School of Mathematics and Statistics, Lanzhou University, Lanzhou 730000, China W. Li (*) School of Mathematics and Statistics, Center for Data Science, Laboratory of Applied Mathematics and Complex Systems, Lanzhou University, Lanzhou 730000, China e-mail: [email protected]
are utilized to model and predict the seasonal subseries and remainder subseries respectively. Finally, two parts are summed to generate the final predictive results. The empirical studies of fossil fuels, nuclear electric power, and renewable energy consumptions show that the proposed model outperforms other alternative benchmarks with regard to effectiveness and scalability. Besides, the sample extrapolation forecasting displays that the technique could limit the error of monthly energy consumption to 3.3%. Keywords Energy consumption prediction . Seasonal adjustment method . Echo state network . Ensemble empirical mode decomposition . Grasshopper optimization algorithm Nomenclature ANN Artificial neural network LSTM Long short-term memory ARIMA Autoregressive integrated moving average MAE Mean absolute error BPNN Back propagation neural network MAPE Mean absolute percentage error CEEMD Complete ensemble empirical mode decomposition MLR multiple linear regression DE Differential evolution MNN Multilayer neural network EEMD Ensemble empirical mode decomposition NMGM Nonlinear metabolic grey model ELM Extreme learning machine NN Neural network
Energy Efficiency (2020) 13:1505–1524
1506
EMD PSO ENN RF ESM RMSE ESN RNN GA SARIMA GM SNN GOA STL GWO SVR IMF WOA LSSVR XG
Data Loading...
