Similar Day Clustering-based Power Load Forecasting For Iron And Steel Enterprises

Steel enterprises as the pillar industry of the national economy,its energy conservation and emission reduction is a major demand for industry development.Among them,electric power as an important secondary energy,the accurate prediction of its load is the key to reduce production cost and improve energy efficiency.However,the production process of steel enterprises is complicated and there are frequent switching conditions,which also increases the difficulty of power load prediction.Aiming at the problem of large fluctuation and poor regularity of power load in steel enterprises,a power load forecasting model based on similar diurnal clustering is established in this paper.Considering the daily similarity of steel production conditions and the limitation of Euclidean distance in measuring the morphological similarity of load data,a column correlation segmentation clustering method based on numerical difference correction was proposed to extract the load characteristics of different similar days,and then the long short term memory(LSTM)neural network was used to model the load of different similar days.In order to improve the prediction accuracy of the model,an adaptive mutation and disturbance acceleration particle swarm optimization(AMDAPSO)algorithm is proposed to determine the model parameters.This method improves the population diversity through the global adaptive probability mutation mechanism,which improves the problem that the conventional particle swarm optimization algorithm is easily trapped in the local optimum,and improves the efficiency of parameter optimization through the local disturbance acceleration strategy.Considering the inaccuracy of single model for load prediction under complex conditions,a combined prediction model based on variational mode decomposition and AMDAPSO optimized LSTM was established.Variational modal decomposition is used to decompose similar daily loads into multiple sub-sequences,and LSTM models are built for them respectively.Then,the proposed AMDAPSO algorithm is used to optimize the parameters of different sub-models,and the load prediction results are obtained by combining them,thus reducing the uncertainty in the process of load prediction.The proposed clustering algorithm,the improved particle swarm optimization process and the combined prediction model were verified and analyzed by using the actual power load data of domestic large iron and steel enterprises.By comparing with the existing algorithms,it is shown that the proposed method is effective for the power load prediction of iron and steel enterprises,and can provide a basis for the production and scheduling of iron and steel enterprises.