国网物资价格预测对预算编制和成本控制至关重要。本文基于国网江苏物资公司历史数据,以TOPSIS法筛选出了五类关键物资,并通过Holt指数平滑法构建了价格预测模型,分析了单期和多期情形下的价格趋势特征。为进一步提升预测精度,利用粒子...国网物资价格预测对预算编制和成本控制至关重要。本文基于国网江苏物资公司历史数据,以TOPSIS法筛选出了五类关键物资,并通过Holt指数平滑法构建了价格预测模型,分析了单期和多期情形下的价格趋势特征。为进一步提升预测精度,利用粒子群算法对模型超参数进行优化,通过与BP神经网络、SVR、Xgboost等模型对比,发现Holt指数平滑法表现出更高预测精度。本研究可为电网企业提供科学决策依据,提升采购合理性和经济性,增强企业市场竞争力与运营稳定性。Price prediction of State Grid materials is critical for budget preparation and cost control. This study utilizes historical data from State Grid Jiangsu Material Company, employing the TOPSIS method to identify five categories of key materials. A price prediction model is constructed using the Holt exponential smoothing method, analyzing price trends and seasonal characteristics under single-period and multi-period scenarios. To improve prediction accuracy, particle swarm optimization (PSO) is applied to optimize model hyperparameters. Comparative analysis with BP neural networks and SVR demonstrates the superior accuracy of the Holt exponential smoothing method. This research provides a scientific basis for power grid enterprises to enhance procurement rationality and economic efficiency, strengthening their market competitiveness and operational stability.展开更多
传统的混凝土拱坝位移预测模型主要关注水压、温度、时效等因素与拱坝位移之间的关系,未对拱坝位移数据中所包含的信息进行充分挖掘。为此,采用Seasonal and Trend decomposition using Loess算法(STL)将拱坝位移原始数据分解为趋势序...传统的混凝土拱坝位移预测模型主要关注水压、温度、时效等因素与拱坝位移之间的关系,未对拱坝位移数据中所包含的信息进行充分挖掘。为此,采用Seasonal and Trend decomposition using Loess算法(STL)将拱坝位移原始数据分解为趋势序列、周期序列及残差分量。在此基础上,采用鲸鱼优化算法(WOA)结合随机森林算法(RF)对三个分量进行预测,并使用Holt-Winters算法充分考虑趋势序列中的趋势信息对趋势序列的预测结果进行修正。最后将修正后的趋势序列预测结果和周期序列、残差分量预测结果相加,得出拱坝位移最终预测结果。工程实例表明,基于STL-Holt-WOA-RF的拱坝位移预测模型能够显著提高预测的准确性和稳定性,为拱坝位移预测提供了新的思路和方法。展开更多
文摘国网物资价格预测对预算编制和成本控制至关重要。本文基于国网江苏物资公司历史数据,以TOPSIS法筛选出了五类关键物资,并通过Holt指数平滑法构建了价格预测模型,分析了单期和多期情形下的价格趋势特征。为进一步提升预测精度,利用粒子群算法对模型超参数进行优化,通过与BP神经网络、SVR、Xgboost等模型对比,发现Holt指数平滑法表现出更高预测精度。本研究可为电网企业提供科学决策依据,提升采购合理性和经济性,增强企业市场竞争力与运营稳定性。Price prediction of State Grid materials is critical for budget preparation and cost control. This study utilizes historical data from State Grid Jiangsu Material Company, employing the TOPSIS method to identify five categories of key materials. A price prediction model is constructed using the Holt exponential smoothing method, analyzing price trends and seasonal characteristics under single-period and multi-period scenarios. To improve prediction accuracy, particle swarm optimization (PSO) is applied to optimize model hyperparameters. Comparative analysis with BP neural networks and SVR demonstrates the superior accuracy of the Holt exponential smoothing method. This research provides a scientific basis for power grid enterprises to enhance procurement rationality and economic efficiency, strengthening their market competitiveness and operational stability.
文摘传统的混凝土拱坝位移预测模型主要关注水压、温度、时效等因素与拱坝位移之间的关系,未对拱坝位移数据中所包含的信息进行充分挖掘。为此,采用Seasonal and Trend decomposition using Loess算法(STL)将拱坝位移原始数据分解为趋势序列、周期序列及残差分量。在此基础上,采用鲸鱼优化算法(WOA)结合随机森林算法(RF)对三个分量进行预测,并使用Holt-Winters算法充分考虑趋势序列中的趋势信息对趋势序列的预测结果进行修正。最后将修正后的趋势序列预测结果和周期序列、残差分量预测结果相加,得出拱坝位移最终预测结果。工程实例表明,基于STL-Holt-WOA-RF的拱坝位移预测模型能够显著提高预测的准确性和稳定性,为拱坝位移预测提供了新的思路和方法。
