三峡大学学报（自然科学版）2025，Vol.47Issue(4)：88-95,8.DOI:10.13393/j.cnki.issn.1672-948X.2025.04.012
贫样本约束下的季度用电量最优组合预测
Quarterly Electricity Consumption Optimal Combination Prediction and Application under Poor Sample Constraints
摘要
Abstract
Reasons such as changes in administrative divisions will limit the amount of electricity used in historical quarters,and the accurate forecasting is challenged.Therefore,an optimal combination prediction model of quarterly electricity consumption under the poor samples constraint is proposed in this paper.Firstly,the quarterly electricity consumption data is decomposed into three types of components:long-term,periodic and random.Moreover,the least squares method and comprehensive accuracy are adopted to divide the components into various components to select the best fitting function.Then,the weighted combination of each fitting function is obtained,the Fruit Fly Optimization Algorithm is used to obtain the optimal combination coefficient,the comprehensive accuracy score is optimized,and the parameters of the model are obtained.A total of 16 sets of quarterly electricity consumption data in the first four years in Chongqing Municipality are collected to verify the model,and the results show that the model can reach 8.962%for fourth-year quarterly electricity consumption prediction index mean absolute percentage error,which is lower than that of the existing four types of prediction models.In addition,the proposed model is applied to Jilin and Fujian,and the results show that the minimum mean absolute percentage error value in the proposed model can reach 2.472%.关键词
季度用电量预测/贫样本约束/趋势分解/综合精度分/果蝇优化算法Key words
quarterly electricity consumption forecasts/poor sample constraints/trend decomposition/comprehensive accuracy score/fruit fly optimization algorithm分类
信息技术与安全科学引用本文复制引用
邓文奇,缪书唯,李振兴..贫样本约束下的季度用电量最优组合预测[J].三峡大学学报（自然科学版）,2025,47(4):88-95,8.基金项目
国家自然科学基金项目(52077120) （52077120）
