基于ISSA-LSTM的储麦长期品质预测
Prediction for Long-Term Quality of Stored Wheat Based on ISSA-LSTM
吴兰 1王恒 1姚远1
作者信息
- 1. 河南工业大学电气工程学院,郑州 450000
- 折叠
摘要
为了解决非时序预测模型无法预测储麦品质时序劣变趋势,以及现有数据驱动的时序预测模型在长期储麦品质预测中因样本不足导致长期预测精度不高的问题,提出一种基于改进麻雀搜索算法(ISSA)优化长短时记忆网络(LSTM)的长期储麦品质预测模型.首先,提出了一种统计均匀分布方法,利用小麦稳定劣化的生理知识对原始数据进行增强扩容.其次,利用麻雀搜索算法(SSA)对LSTM模型进行优化,克服局部极值点,提高收敛速度.最后,引入t分布函数对SSA位置更新过程进行扰动避免局部最优.结果表明,储麦品质参数中的吸水率、咀嚼度、脂肪酸值和峰值黏度与储藏时间的Spearman相关性较为显著,相关系数均高于0.9,ISSA-LSTM模型预测精度相比于BP、LSTM、SSA-LSTM预测模型分别提高了11.83％、16.98％、26.50％,有助于提高小麦品质预测及分析的准确性.
Abstract
The non-time series prediction model cannot predict the trend of time series deterioration of stored wheat quality,while existing data-driven time series prediction model has low long-term prediction accuracy due to insufficient samples in long-term stored wheat quality prediction.In this way,a long-term quality prediction model of stored wheat based on improved Sparrow Search Algorithm(ISSA)optimized Long Short-Term Memory Network(LSTM)was proposed.Firstly,a statistical uniform distribution method was proposed to enhance and expand the o-riginal data by using the physiological knowledge of stable deterioration of wheat.Secondly,the Sparrow Search Algo-rithm(SSA)was used to optimize the LSTM model to overcome the local extreme points and improve the convergence speed.Finally,the T-distribution function was introduced to perturb the SSA position update process to avoid local optima.The results indicated that the Spearman correlation between water absorption,chewiness,fatty acid value and peak viscosity and storage time was significant,and the correlation coefficients were higher than 0.9.Compared with BP,LSTM and SSA-LSTM prediction models,the prediction accuracy of ISSA-LSTM model was improved by 11.83％,16.98％and 26.50％,respectively,helpful to improve the accuracy of wheat quality prediction and analysis.
关键词
模式识别与智能系统/储藏小麦品质/预测模型/长短时记忆网络/麻雀搜索算法/统计均匀分布Key words
pattern recognition and intelligent system/stored wheat quality/prediction model/long short-term memory network/sparrow search algorithm/statistical uniform distribution引用本文复制引用
基金项目
国家自然科学基金项目(61973103)
河南省高校科技创新团队项目(24IRTSTHN030)
河南省科技厅自然科学项目(222102220009)
郑州市科技局自然科学项目(22ZZRDZX06)
出版年
2024
NETL
NSTL
万方数据