基于组合模型的天然气管道短期负荷预测
Short-term Load Forecasting of Natural Gas Pipeline Based on Combination Model
陈宇光1
作者信息
- 1. 中国石油天然气股份有限公司新疆油田分公司吉庆油田作业区,新疆吉木萨尔 831700
- 折叠
摘要
通过灰色关联分析(GRA)确定影响天然气管道短期负荷的主控因素,应用非线性变化策略对粒子群(PSO)算法的惯性权重和加速因子改进,利用优化算法寻找适合长短期记忆网络(LSTM)模型的超参数,形成GRA-IPSO-LSTM的组合模型,并与其余模型对比,以验证其准确性和可靠性.结果表明:根据灰色关联度的大小,可以逐步删减对日负荷影响不大的因素,降低后续预测模型的复杂性;IPSO算法在迭代速度、收敛精度和寻优质量方面均有所提高,降低了 LSTM模型超参数人工选取的局限性;该组合模型的平均绝对百分比误差(MAPE)和均方根误差(RMSE)均得到大幅降低,决定系数(R2)和预测精度得到提高,证明了该组合模型可用于天然气管道短期负荷的准确预测.
Abstract
Grey correlation analysis(GRA)is used to determine the main control factors affecting the short-term load of natural gas pipeline.Nonlinear change strategy is applied to improve the inertia weight and acceleration factor of particle swarm optimization(PSO)algorithm.The optimization algorithm is used to find the hyperparameters suitable for the long short-term memory network(LSTM)model,and a combination model of GRA-IPSO-LSTM is formed.It is compared with other models to verify its accuracy and reliability.The results show,according to the size of grey correlation degree,the factors that have little influence on daily load can be deleted step by step,the complexity of the subsequent prediction model can be reduced.iteration speed,convergence accuracy and optimization quality are improved in IPSO algorithm.The limitation of manual selection of LSTM model hyperparameters is reduced.The MAPE and RMSE of the combined model are sharply reduced,theR2 and the prediction accuracy are improved.It proves that the combined model can be used to accurately predict the short-term load of natural gas pipelines.
关键词
短期负荷预测/灰色关联分析/粒子群算法/长短期记忆网络Key words
short-term load forecasting/grey correlation analysis/particle swarm optimization/long short-term memory network引用本文复制引用
出版年
2024
NETL
NSTL
万方数据