English

Journal of Chinese Agricultural Mechanization

Journal of Chinese Agricultural Mechanization ›› 2024, Vol. 45 ›› Issue (1): 169-176.DOI: 10.13733/j.jcam.issn.2095-5553.2024.01.024

Previous Articles     Next Articles

Prediction of environmental parameters and early warning of cherry cracking in greenhouse by GAAttentionLSTM algorithm

Hu Lingyan1, Qiu Shaohang1, Li Guoqiang2, Xu Wei1, Liu Yan1, Wang Zumin1   

  • Online:2024-01-15 Published:2024-02-06

融合GA-Attention-LSTM算法的温室樱桃环境参数预测与裂果预警

胡玲艳1,邱绍航1,李国强2,许巍1,刘艳1,汪祖民1   

  • 基金资助:
    国家自然科学青年基金(61601076);大连市科技创新基金项目(2020JJ26SN058、2021JJ13SN78)

Abstract: Aiming at the influence of greenhouse environmental factors on cherries, a set of automatic environmental monitoring device for large cherry greenhouse was designed to collect environmental parameter values in the greenhouse to provide digital early warning support and control plan for cherry cracked fruit. Based on the environmental parameter values, the correlation analysis was used to obtain the environmental parameter characteristics with strong correlation with the cracked fruit in the shed. Secondly, the sliding window method was used to generate the input environment features into a time series matrix form. Then, a prediction model integrating the GAAttentionLSTM algorithm was proposed to accurately predict the environmental parameters in the shed. Finally, SPSS data analysis software was used to analyze the environmental parameters and fruit splitting rate of different greenhouses. The average absolute error of the proposed prediction model with GAAttentionLSTM algorithm was 0.112 and the mean squared error was 0.087, which was 12.80% and 9.72% higher than that of the LSTM network model, and the prediction accuracy of environmental parameters was higher, and a set of scientific cherry environmental parameter value ranges was obtained, which could provide strong support for the prediction model for the digital early warning of cherry split fruit.

Key words: smart agriculture, greenhouse cherry, LSTM model, environmental parameter, fruit cracking warning, accurate prediction

摘要: 针对温室环境因素对樱桃的影响,设计一套大樱桃温室环境自动监测装置,用来采集温室内的环境参数值为樱桃裂果提供数字化预警支持及防治方案。基于采集的环境参数值,首先使用相关性分析得出与棚内裂果具有强相关性的环境参数特征;其次使用滑动窗口方法将输入的环境特征生成时间序列矩阵形式;随后提出一种融合GAAttentionLSTM算法的预测模型,实现精准预测棚内的环境参数的功能;最后通过SPSS数据分析软件来分析不同大棚的环境参数和裂果率。所提的融合GAAttentionLSTM算法的预测模型的平均绝对误差为0.112,均方误差为0.087,相比于LSTM网络模型高出12.80%和9.72%,对环境参数的预测精度更高,同时得出一套科学的樱桃环境参数值范围,为预测模型对樱桃裂果数字化预警提供有力支持。

关键词: 智慧农业, 温室樱桃, LSTM模型, 环境参数, 裂果预警, 精准预测

CLC Number: