Probabilistic short-term power load forecasting based on B-SCN

Grid management and power dispatching rely on accurate short-term power load prediction. Different algorithms have been constantly developed and tested to improve forecast precision. However, these forecasts are constrained by a number of uncertain factors, which are caused by dynamic environment, the nonlinearity and stochasticity of power demand. To obtain more accurate load forecasting value and quantify the uncertainty effectively, this research proposes a boosting stochastic configuration network(B-SCN) based probabilistic forecasting method. First, correlation analysis is taken in multidimensional input parameters. Second, an adaptive B-SCN network architecture is proposed to construct the prediction model and improve the stability of model outputs significantly. The probabilistic forecasting is then used to actualize the model's uncertainty evaluation by creating the confidence intervals using the Gaussian process. Consequently, experimental results reveal that the proposed boosting-SCN prediction model achieves superior forecasting accuracy than the single SCN model and other commonly used forecasting models. The probabilistic forecasting can efficiently obtain the uncertainties in power load data and provide support for system operation.(c) 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).