A sparse heteroscedastic model for the probabilistic load forecasting in energy-intensive enterprises

The energy-intensive enterprises (EIEs) account for a significant part of the total electricity consumption in most industrial countries. In the smart grid environment, electric load forecasting in EIEs plays a critical role in the security and economical operation of both the main grid and the EIEs' micro-grid. However, the accuracy of such forecasting is highly variable due to the strong stochastic nature of the load in EIEs. In this circumstance, probabilistic forecasts are essential for quantifying the uncertainties associated with the load, thus is highly meaningful for assessing the risk of relying on the forecasts and optimizing the energy systems within EIEs. This paper focuses on the day-ahead probabilistic load forecasting in EIEs, a novel sparse heteroscedastic forecasting model based on Gaussian process is developed. With the proposed model, we can provide predictive distributions that capture the heteroscedasticity of the load in EIEs. Since the high computational complexity of Gaussian process hinder its practical application to large-scale problems such as load forecast, the proposed model employs the ℓ_1/2 regularizer to reduce its computational complexity, thereby enhancing its practical applicability. The simulation on real world data validates the effectiveness of the proposed model. The data used in the simulation are obtained in the real operation of an EIE in China.