Forecasting the electric power load based on a novel prediction model coupled with accumulative time-delay effects and periodic fluctuation characteristics

The electric power load forecasting has important theoretical and practical significances for balancing electric power supply. To produce the precise values of the future electric power load, a new discrete grey multivariable prediction model with accumulative time-delay effects and periodic fluctuation characteristics is proposed. In the new model, the accumulative time-delay effects and periodic fluctuation are analyzed and represented by the algebraic expressions. The discrete expression of the new model, containing both the accumulative time-delay effects and periodic fluctuation characteristics, is put forward. A theorem is developed to give the matrix representations of the unknown parameters through the least squares estimation. Then, the vector autoregression model is carried out to identify the accumulative time-delay parameters. A nonlinear programming problem is developed to optimize the parameters by using the whale optimization algorithm. Finally, the model is implemented to predict the monthly electricity consumptions in Zhejiang Province to validate its advantages. Results show the new model has better performances than benchmark models. The model decreases the MAPE value by 121.05 %, 32.58 %, 238.35 % and 189.47 % than similar grey models and reduces the MAPE value by 20.05 %, 17.54 %, 285.21 % and 275.69 % than the statistics method models and the machine leaning models.