Very short-term demand forecasting for fused magnesium furnaces based on specific penalty function

As the manufacturing equipment of the refractory-fused magnesia, fused magnesium furnace (FMF) has the following characteristics: huge power consumption, peak load impact and strong fluctuations. During the smelting process, the total load of the FMF group often produce electricity demand spikes, which are unfavorable to the production. More seriously, the demand may exceed the contract value and this case will affect the safe operation of the furnace and grid. As the randomness of the furnace, in different working conditions, increases the uncertainty of the demand of FMF group, accurate demand forecasting is very important for monitoring the demand trends. In order to meet the asymmetric requirements for forecasting error, this paper presents a asymmetrical penalty function, uses a combination of methods BAT-RBF to design the predictor, in which the BAT algorithm are used to optimize the weight coefficients of the RBF network. In addition, some experiments based on the field data of Liaoning, by comparison with other methods, demonstrates the effectiveness and feasibility of the proposed method.