基于IPSO算法优化GRU神经网络的燃气负荷预测  

Gas load forecasting based on optimization of GRU neural network using IPSO algorithm

作　　者：海文龙 王亚慧 王怀秀 HAI Wenlong;WANG Yahui;WANG Huaixiu(Department of Electrical and Information Engineering,Beijing University of Civil Engineering and Architecture,Beijing 100044,China)

机构地区：[1]北京建筑大学电气与信息工程学院,北京100044

出　　处：《传感器与微系统》2023年第2期126-129,共4页Transducer and Microsystem Technologies

基　　金：国家重点研发计划支撑项目(2018YFC0807806)。

摘　　要：燃气负荷具有非平稳性、非线性等特点,机理建模困难。为寻求更准确的燃气负荷预测方法,提出一种基于改进粒子群(IPSO)算法优化门控循环单元(GRU)的燃气负荷预测模型。首先,确定了模型的结构与输入变量;然后,针对长短期记忆(LSTM)神经网络复杂度高,训练速度慢的问题,提出采用相较于LSTM单元结构更加简单的GRU作为预测模型,并采用粒子群优化(PSO)算法对GRU网络参数进行优化,引入非线性惯性权重对PSO算法进行改进。实验表明:所提模型相对误差为1%,精度高于传统LSTM与GRU模型。Gas load is non-stationary and non-linear, so it is difficult to model the mechanism.In order to find a more accurate gas load forecasting method, a gas load forecasting model based on improved particle swarm optimization(IPSO)is proposed to optimize gated recurrent unit(GRU).Firstly, the structure and input variables of the model are determined, and then for the problems of high complexity and slow training speed of the long short-term memory(LSTM)neural network, GRU,which has a simpler structure compared with the LSTM unit, is proposed as the prediction model, and the PSO algorithm is used to optimize the GRU network parameters.The PSO algorithm is improved by introducing nonlinear inertia weights.The experiments show that the relative error of the established model is 1 %,its precision is higher than traditional LSTM and GRU models.

关 键 词：燃气负荷预测 粒子群优化算法 长短期记忆神经网络 非线性惯性权重 门控循环单元 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TU996[自动化与计算机技术—计算机科学与技术]