Research On Prediction Technology Of Temperature And Humidity In Solar Greenhouse Based On LSTM Model

With the implementation of the rural revitalization strategy,solar greenhouse planting has become the main pillar industry to increase farmers’ income and become rich.Compared with the field planting environment,the advantage of the solar greenhouse is that it can effectively resist the invasion of natural disasters such as high temperature,low humidity and frost,so that the crops can be in a suitable growth environment for a long time.Temperature and humidity are important environmental parameters that affect crop growth.If The temperature and humidity are not regulated in time in a relatively closed greenhouse environment,it is easy to form environmental conditions such as high temperature and high humidity that are not conducive to crop growth.However,the environmental regulation of the greenhouse usually occurs after the situation that harms the growth of crops has been formed,and then the management personnel implement the corresponding regulation strategy to improve.There is a serious lag in this kind of greenhouse environmental regulation,which is easy to cause plant diseases and insect pests,resulting in crop yield reduction and economic losses.The reason for such consequences is that the regulation strategy lacks the prediction of the temperature and humidity in the greenhouse.Therefore,this paper studies the temperature and humidity prediction technology of the solar greenhouse.Provide solutions to solve the lagging problem of environmental regulation strategies.The main contents of this article are as follows:1.In order to solve the problem that there are many environmental parameters affecting the temperature and humidity change and the coupling relationship is complex,we analyzed the correlation between the collected environmental parameters inside and outside the greenhouse and temperature and humidity based on Pearson correlation theory.The study found that among the environmental parameters such as light intensity,solar radiation,carbon dioxide,air temperature,humidity,soil temperature and soil humidity,only soil temperature was not correlated with temperature and humidity,and soil humidity was correlated with relative humidity.Then we carried out the autumn temperature and humidity prediction experiments on the existing three models.The experimental results show that the RMSE value of the GRU model is 1.27 ℃ in terms of temperature prediction,which is the best among the three models.GRU is 0.23℃ and 0.27℃ lower than the LSTM and ANN models,respectively,and the GRU and LSTM models both have 98%accuracy in predicting temperature.In terms of humidity prediction,the LSTM model is better than the other two models.The LSTM model predicts the humidity MAE value of 1.5%RH,which is lower than the GRU and ANN models by 0.27%RH and 0.54%RH.2.In view of the lack of consideration of the correlation between temperature and humidity and other input parameters in the existing model for temperature and humidity prediction,which leads to inaccurate temperature and humidity prediction,In this paper,a Temperature and Humidity Attention Mechanism-LSTM(THA-LSTM)model for solar greenhouse is designed.The attentional mechanism helps the model to allocate corresponding weight to the output sequence of hidden layer,so that the model pays more attention to the parameters with strong correlation affecting the prediction target,and improves the accuracy of temperature and humidity prediction of the model.Two sets of meteorological stations inside and outside the greenhouse were used to collect data.After data pretreatment,data sets were divided into autumn and winter to form data sets,and prediction and comparison experiments were conducted on the autumn and winter data sets in the greenhouse.In the indoor temperature and humidity prediction experiment in autumn and winter,the temperature prediction error of THA-LSTM model is 35%and 28%lower than MAE and RMSE value of original LSTM model.In terms of humidity prediction,MAE value of THA-LSTM model was 1.36%RH,which was 0.14%RH,0.41%RH and 0.68%RH lower than LSTM,GRU and ANN models.The experimental results show that MAE and RMSE of THA-LSTM model are 0.67℃ and 1.25℃ for indoor temperature and humidity prediction by outdoor environmental meteorological data set in autumn and winter;In terms of humidity prediction,MAE and RMSE values of THALSTM model are 1.61%RH and 2.54%RH,which are 39%and 19%lower than those of original LSTM model.3.Develop a solar greenhouse environment monitoring and prediction system based on Raspberry Pi.First,the sensors monitoring environmental parameters are connected to the concentrator to build a meteorological monitoring station inside and outside the greenhouse.The software system runs on the Raspberry Pi,and then the system collects and displays the environmental monitoring data in real time.Finally,according to the saved data,the THA-LSTM model is used to predict the temperature and humidity in the next three days.change and visualize.