Multivariate statistical and similarity measure based semiparametric modeling of the probability distribution: A novel approach to the case study of mid-long term electricity consumption forecasting in China

To achieve the goal of drawing up optimal plans for power generation, decision makers need an appropriate methodology to effectively identify the pivotal aspects of electricity consumption fluctuation and anticipate the future trend. The parameter identification of conventional statistical approach mainly relies on distributional assumptions and functional form restrictions, which might be problematic for the real application. This paper addresses these issues by implementing a novel semi-parametric modeling approach, which is suitable for investigating the uncertainties in the mid-long term forecast and estimating the probability distributions of future demand. To identify the significant impact factors of the electricity consumption, a new Kullback-Liebler (K-L) divergence based similarity measure strategy is designed. A case study concerning the electricity demand forecasting in China demonstrates the applicability of the proposed approach and verifies the feasibility of establishing explicit functional dependency between external variables and electricity consumption. Despite the complexity, notable reductions in the number of forecasting error are obtained due to the adoption of three indicators: deposits in financial institutions, exports, and imports.