Long Term Electricity Consumption Forecast Based on STL-IDBO-LSSVM

Electricity Consumption is used to measure the level of economic development of a region, which is a authoritative symbol. In this study, a seasonal and trend decomposition using loess (STL) and a improved dung beetle optimization (IDBO) were embedded as compositional algorithm in least squares support vector machine (LSSVM) algorithms, with the aim of forecasting long term power. Firstly, the drawback of the Dung Beetle Algorithm (DBO) is that it has an unbalanced ability to explore globally and exploit locally, so an IDBO is proposed based on variation of self-adaptive t distribution, which is more superior compared with Sparrow Search Algorithm (SSA), DBO and Particle Swarm Optimization (PSO), then builds the IDBO-LSSVM power prediction model. Secondly, with the purpose of fully exploiting the features of the power series, the STL algorithm splits the power into a trend component, a seasonal component and a residual component, which reflect the long term trend and periodical changes of power data, each time series components are used as inputs to the IDBO-LSSVM prediction model. Finally, the results indicated that the method has a prediction accuracy of 97.31% through experimental comparisons, which is better than the other five models, proving the feasibility and superiority of the proposed method.